d6.1 design specification v1.0 - commrob - home · figure 23: scheme for rfid based navigation in...

D6.1 – Design specification

CommRob IST-045441 Advanced Behaviour and High-Level Multimodal Communication with and among Robots

page 1/54

D6.1 – Design specification Workpackage WP6 Document type Technical Document Title D6.1 – Design specification Authors ETP (WP6-Leader) Internal Reviewers FZI, LAAS, TUW, ZENON Version V 1.0 Status Final Distribution Public



page 2/54

History of changes

Date Ver. Author(s) Change description

10.01.08 0.1 Lutz Kettenhofen (ETP) Initial document

15.01.08 0.2 Lutz Kettenhofen (ETP) Internal review

22.01.08 0.3 Thomas Weeber, Eckhard Gerland; Lebrecht von Blücher (ETP)

Internal review

29.01.08 0.4 George Nikolakopoulos (ZENON)

Contribution from ZENON

30.01.08 0.5 Lutz Kettenhofen Contribution from ETP

30.01.08 0.6 George Nikolakopoulos Comments from ZENON

30.01.08 0.7 Thomas Weeber; Lutz Kettenhofen (ETP)

Internal review


Implementing revised contributions from TUW


Implementing revised contributions from LAAS


Implementing revised contributions from FZI

15.02.08 0.11 George Nikolakopoulos, Vasilis Spais (ZENON)

Revised Contribution ZENON


Internal review

28.02.08 0.13 Lutz Kettenhofen Improvement of figures by input from TUW

29.02.08 1.0 Lutz Kettenhofen Internal review



page 3/54

Summary This document contains the detailed design and layout for a demonstration robot of the CommRob project. The document consists of three parts with the following content:

The chapter “Trolley system” explains the deduction of the physical system architecture from the logical view that has been developed in work package WP2. It explains the overall set-up of the trolleys and shows its main components.

The chapter “Vehicle set-up demonstrator” describes the trolley from a mechanical point of view and describes its structures and components.

The chapter “Trolley computer based systems” defines the processing units which are the platforms for the implementation of the software developed in the other work packages.

The chapter “Logical Sensors” contains the specification of the used sensors and sensor systems of the trolley.



page 4/54

Table of contents 1. Introduction.................................................................................................................10

1.1 Purpose of the Document ...........................................................................................10

1.2 Relationship to Other Documents...............................................................................10

2. Trolley system ............................................................................................................11

3. Vehicle Set-Up demonstrator .....................................................................................14

3.1 Bodywork/geometry ....................................................................................................14

3.1.1 Chassis Requirements................................................................................................14

3.1.2 Manufacturing the trolley ............................................................................................18

3.2 Undercarriage, steering and drive system ..................................................................20

3.2.1 Omnidrive Rotatable Wheels ......................................................................................20

3.2.2 Specifications..............................................................................................................22

3.3 Energy supply .............................................................................................................23

3.4 Warning and Safety devices .......................................................................................24

3.4.1 Person detection system ............................................................................................24

3.4.2 Braking system ...........................................................................................................25

3.4.3 Warning devices .........................................................................................................26

3.4.4 Emergency stop..........................................................................................................26

3.5 Operating elements ....................................................................................................26

3.5.1 Handle User Interface.................................................................................................26

3.5.2 Human IN/OUT...........................................................................................................28

4. Trolley computer based systems................................................................................30

4.1 PC for communication ................................................................................................30

4.2 PC for planning and interaction ..................................................................................30

4.3 PC for Execution.........................................................................................................31

4.4 Motion Control System ...............................................................................................31

4.5 Wireless Communication ............................................................................................33

5. Logical sensors...........................................................................................................34



page 5/54

5.1 Audio In.......................................................................................................................34

5.2 Audio Out....................................................................................................................35

5.3 RFID ...........................................................................................................................35

5.3.1 RFID for user identification .........................................................................................35

5.3.2 RFID for navigation.....................................................................................................36

5.4 Wireless ......................................................................................................................39

5.5 Visual Features...........................................................................................................46

5.5.1 Stereo camera for the vision on the user....................................................................46

5.5.2 Camera for the robot localization................................................................................47

5.5.3 Camera for obstacle detection....................................................................................48

5.6 Haptics........................................................................................................................49

5.6.1 Mechanical design of the haptic handle......................................................................49

5.6.2 Electronic components of the haptic handle ...............................................................50

5.6.3 Prototype and first tests of the haptic handle..............................................................50

5.7 Distance......................................................................................................................51

5.7.1 Device for security ......................................................................................................51

5.7.2 Device for obstacle detection......................................................................................52

6. Conclusions ................................................................................................................53

7. References .................................................................................................................54



page 6/54

List of abbreviations

CAN Controller Area Network CSS Chirp Spread Spectrum GUI Graphical User Interface HUI Handle User Interface LBS Location Based Services PID Proportional Integral Derivative PTZ Pan Tilt Zoom RFID Radio Frequency Identification Device RSS Received Signal Strength RSSI Received Signal Strength Indicator RTLS Real Time Location Systems SDS Symmetrical Double-Sided SoC System-on-Chip TDC Time-to-Digital Converter TOF Time-Of-Flight TRX Transceiver TWR Two-Way Ranging UCoM Universal Controller Module UML Unified Modelling Language UWB Ultra Wideband WSNs Wireless Sensor Networks



page 7/54

List of figures

Figure 1: Logical view of system architecture ........................................................................11

Figure 2: UML representation of the deployment ...................................................................12

Figure 3: CommRob Trolley - Equipment indication...............................................................13

Figure 4: CommRob Trolley design, side view.......................................................................14

Figure 5: CommRob Trolley design, front view ......................................................................15

Figure 6: CommRob Trolley design, reverse view .................................................................16

Figure 7: CommRob Trolley design, top view ........................................................................17

Figure 8: CommRob Trolley – Concept Design with Equipment Installed..............................18

Figure 9: Examples of Welded Frames ..................................................................................19

Figure 10: A Type of Frame Built by Sheet Metal ..................................................................19

Figure 11: Cases of Wheels orientation towards direction of moving ....................................21

Figure 12: Rotary Wheels Mechanism – Concept Design......................................................22

Figure 13: Integrated Controller Board...................................................................................23

Figure 14: Energy Supply Unit ...............................................................................................24

Figure 15: Automated guided vehicle with laser scanner.......................................................25

Figure 16: Haptic handle concept drawing .............................................................................26

Figure 17: Modes of driving postures .....................................................................................27

Figure 18: The touch screen panel – Human In\Out interface ...............................................29

Figure 19: Motion Control System..........................................................................................32

Figure 20: Motion Control System physical mapping .............................................................32

Figure 21: The array of 8 microphones developed at LAAS-CNRS. ......................................34

Figure 22: RFID setup mounted on a platform at LAAS-CNRS .............................................35

Figure 23: Scheme for RFID based navigation in CommRob ................................................36

Figure 24: RFID measurement coil ........................................................................................38

Figure 25: RFID measurement area.......................................................................................38

Figure 26: Reading performance of a “double” RFID tag .......................................................38

Figure 27: Distance for placement of RFID tags ....................................................................39



page 8/54

Figure 28: Concept Application of the WSN for Indoor Localization ......................................40

Figure 29: The Chipcon mote.................................................................................................41

Figure 30: The nanoLOC TRX Transceiver............................................................................42

Figure 31: The MICA2 Cricket Location Mote ........................................................................44

Figure 32: Stereo and range imaging camera setup at LAAS-CNRS ....................................47

Figure 33 (left) CAD drawing of the haptic handle (right) Placement of the strain gages on the different strain gauges beams (B1-B4)...................................................................................49

Figure 34: Schematic diagram on the electronic components of the haptic handle ...............50

Figure 35: (left) Prototype of the haptic handle (right) Position of the strain gauges with measurement electronics .......................................................................................................51

Figure 36: Laser scanner Sick S300 Professional .................................................................51



page 9/54

List of tables Table 1: Characteristics of the auditory sensor. .....................................................................34

Table 2: Example of RFID device from LAAS-CNRS.............................................................35

Table 3: Characteristics of the RFID-Reader .........................................................................37

Table 4: Characteristics of the Pan Tilt platform ....................................................................46

Table 5: Characteristics of the range imaging camera used at LAAS-CNRS ........................47

Table 6: Characteristics of the laser scanner Sick S300........................................................52



page 10/54

1. Introduction

1.1 Purpose of the Document The main purpose of this deliverable is to define the detailed design and layout for a demonstration robot. The set-up of the demonstration robot needs to be defined by its mechanical parts and electronics.

The partners have specified these components and the overall set-up of the robot according to the special demands and requirements as defined in their tasks. Therefore the demonstration robot will be able to fulfil the given goals. Special attention has been given to ergonomics and the real supermarket environment of a trolley.

The design specification constitutes the base for the later integration work of the modules as provided by the partners.

1.2 Relationship to Other Documents The design specification builds on the work achieved in the other work packages.

• D2.1 (First Version of System, Architectural and Subsystem specification)

• D3.1 (Report on Behaviour and Navigation Specification and Architecture)

• D4.1 (Report on human-robot interaction specification)

• D4.2.1 (Intermediate report on haptic-based detection)

• D4.3.1 (Intermediate report on basic functions for detection/recognition of people)

• D5.1 (Report on the communication model)

• D5.2 (Requirements definition of communication platform)

• D7.1 (Report on user and stakeholder requirements)



page 11/54

2. Trolley system The design specification of the shopping trolley is based on the system architecture developed by the partners in the CommRob project. This architecture represents the system functionality by partitioning it into well connected components. The aim of this deliverable is to translate and map the requirements and the system architecture from WP2 (and WP3, 4, 5, and 7) into a physical architecture where all the technologies, the components and the outcome of the project will be evaluated and demonstrated.

As it is described in deliverable D2.1 [1], the consortium concluded in a layered architecture whose logical view is shown in figure 1.

Trolley

Human Computer In /Out

HapticHandle

AudioOut

OpticalIn

WirelessCommunication

Wireless UserIdentification

DistanceSensors

OpticalOut

AudioIn

Logical Sensors

«layer»Executive

:Local Nav igation and Motion

«layer»Operations

«layer»Communication

:High-lev el Nav igation

:Speech and Gesture Interpretation

:Communication

:Application Logic

World Model Repository

:User Perception, Identification and Tracking

:Local Env ironment Perception and Object Analysis

:Exteroceptiv e (v ision and haptic based )

Serv oing

:Scene Analysis

:Visual Features

:Audio OUT

:Haptics

:Distance

:RFID/Wireless

:Local World Model Repository

:Ontology Repository

:Audio IN

:Topological Map Repository

«use»

«use»

Figure 1: Logical view of system architecture

For realizing this architecture on computer based platforms, the logical components were grouped into three systems which are connected by a common communication channel (Ethernet).

Each of these computer based systems has its own designation and deals with specific tasks which lead to the following names of the systems:

• PC for communication

• PC for planning and interaction



page 12/54

• PC for execution

A general approach for optimizing processing speed and data handling was taken by assigning specific logical sensors directly to that computer based system that will use them mostly. This will help to reduce latency and improve reaction rate to occurring events.

The handling of the wheel drive will be done by a designated hardware platform named Motion Control System. This embedded system will be connected via CAN bus to the PC for execution on which the related module Local Navigation and Motion is running.

Figure 2 shows a UML representation of the system architecture with the Logical systems assigned to the related system.

«device»PC for planning and interaction

«device»PC for Execution

«device»PC for communication

Ethernet

«device»DSP/MotioncontrolSystem

CAN

:Communication :Application Logic

:Scene Analysis

:High-lev el Nav igation

:Speech and Gesture Interpretation

:User Perception, Identification and

Tracking

:Audio IN

:Audio OUT

:Visual Features

:RFID/Wireless

:Distance

:Haptics

:Local Env ironment Perception and Object Analysis

:Exteroceptiv e (v ision and haptic based )

Serv oing:Local Nav igation and Motion

:Local Nav igation and Motion

:Local World Model Repository

:Ontology Repository

:Topological Map Repository

Figure 2: UML representation of the deployment

Additionally the overall system should allow communication from trolley to the maintenance station and communication between trolleys by the usage of a wireless communication system (wireless Ethernet).



page 13/54

Based on this logical view, the partners defined in WP6 which parts and components will be used for the robot trolley. The overall bodywork was created with respect to the requirements of the partners, so that all modules, bought from the shelf or developed in other WPs, can be integrated.

Figure 3 shows the trolley construction with the specific hardware and the computer based systems as projected for CommRob. Every component is identified and named in the figure. Detailed information like the type and functionality of the components are specified in the following chapters.

For giving help in finding these components, the figure shows the name of the components as well as the chapter number, where it can be found. So the detailed specification of e.g. Human IN/OUT can be found in chapter 3.5.2, as indicated by “s. 3.5.2”.

Figure 3: CommRob Trolley - Equipment indication



page 14/54

3. Vehicle Set-Up demonstrator

3.1 Bodywork/geometry

3.1.1 Chassis Requirements The CommRob trolley geometry was influenced by the shape of common shopping trolleys. However, major modifications have been done to be able to carry the equipment needed for the operation of the CommRob system.

This equipment includes communication devices, human-machine interface, motion control equipment and in general all the hardware modules which are presented in Figure 3. Furthermore, there is a basket for collecting goods as this is an essential characteristic of a robot to be used as a supermarket trolley.

Another thing of paramount importance is the reservation of an adequate space to install the power supply and the electrical cabinets. Cable routing is also another issue that was taken under very serious consideration as the needs and the numbers of the necessary cables are complicating the mechanical design of the CommRob trolley. On the other hand, having provision for supporting the cables on the chassis, makes the assembling procedure easier and can release space that otherwise would be occupied by cables.

After an extensive analysis of the specification and the functionalities that the autonomous CommRob trolley should satisfy the partners concluded in a first realistic design concept, which is presented in the Figures 4, 5,6,and 7as it is seen from various orthogonal angles.

Figure 4: CommRob Trolley design, side view



page 15/54

It is of high importance to note that this design concept is the result of an iteration phase among the partners that have lasted over six months. This design concept has been made targeting the real manufacturing of the trolley. In the evolvement of this deliverable specific mechanical and electrical drawings will be produced.

As this task is an ongoing task, further analysis and evaluations on the advantages and disadvantages of this concept will be performed from all the partners of the consortium.

Figure 5: CommRob Trolley design, front view

In order to have an understanding of the actual physical size of the trolley, the interested reader should refer to Figure 4, where it is shown that the trolley has a total height of 1.9m,



page 16/54

the rod for supporting the stereoscopic camera included. If we only take the distance from the handles to the ground, the height is 1.15m, comparable to the trolleys used in today’s markets. The trolley has a length of 1.7m which is widely used for supermarket trolleys, too.

Figure 6: CommRob Trolley design, reverse view



page 17/54

Figure 7: CommRob Trolley design, top view



page 18/54

3.1.2 Manufacturing the trolley There are two possibilities to build the trolley chassis. It can either be made by welded structural parts or by bended sheet metal plates. A general view of the trolley with all the necessary equipment installed is presented in Figure 8. This trolley has been designed in order to be able to carry a payload of 100kg and to fulfil the specifications of speed and range (Table 1):

Specification Value

Maximum Cruising Speed 2m/sec

Maximum Sensors’ Range 500cm

Table 1: Specification of speed and range

Figure 8: CommRob Trolley – Concept Design with Equipment Installed



page 19/54

At the first option the result is a very stiff frame made by structural parts of circular, rectangular or angle type sections. There are a few options regarding the material of these parts that can affect the overall weight of the structure. Steel structural profiles result in a simple solution that can be joined easily by welding. On the other hand, aluminium structural parts result in a light structure, however the costs are higher for purchasing these components and for joining them together as well.

The concept of a welded frame is similar to those used for bicycles or small vehicles. The main advantage is the increased stability. Depending on the material the dimensions of the structural parts, the desired stiffness, and the weight of the frame may vary.

Figure 9: Examples of Welded Frames

The frame can be covered by various materials to make the trolley look better and to cover the interior components. In the concept presented in figure 10, the frame is covered by plastic sheets properly shaped by vacuum forming to the desired geometry. Sheet metal plates could also be used if it is possible to form them to the desired geometry.

As an alternative to produce CommRob trolley chassis based on tubes of ring or rectangular cross sections, bended sheet metal parts could be used.

Figure 10: A Type of Frame Built by Sheet Metal

The weight of a sheet metal chassis can be reduced in comparison to a steel tube welded frame. However, it cannot be as stiff as a welded frame. Maximum supported weight is also limited by the strength of each sheet metal part. This value is possible to be increased by adding stiffening parts at specific positions where the load is applied.



page 20/54

In both cases, this structure can be made using bolts instead of welding.

At specific points of the frame, there are glands or connectors to allow cables enter the chassis. They are routed next to the frame structural parts and come out at the desired points (e.g. near the electric device).

There are a few points, like the mounting flanges of the rotatable wheels that need further support in both cases especially at the metal sheet type of chassis which is less stiff. Adding stiffening parts can increase the endurance of the trolley and allow more accurate control of its motion.

3.2 Undercarriage, steering and drive system

3.2.1 Omnidrive Rotatable Wheels The selection of the omnidrive rotatable wheels is mandatory for CommRob as the ability of the trolley to manoeuvre in a minimum of space is of paramount importance for having a vehicle which would look like a normal super-market trolley.

The manoeuvrability of the trolley will be achieved through omnidrive rotatable wheels. These omnidrive wheels have become most popular in robotic applications as they allow the robotic platform to steer in all necessary directions. In this four wheels drive it is possible for all the wheels to rotate separately and thus one can go straight, in an angle or rotate. The selection of the omnidrive wheels should be made most carefully considering the advantages and disadvantages of each selected technology.

For an example in case of four stand-alone rotatable wheels, For achieving the required movement, the wheels have to be turned in place before a direction change is possible. At sharp direction changes (this is a usual case for the CommRob trolley) this operation might take some minimum time upon completion and accordingly this should be taken under consideration in the motion planning algorithm. Some typical cases of wheels orientation with respect to the motion required are explained and displayed in the following cases in Figure 11.

Case 1: Moving Straight Ahead

All wheels rotate at the same speed and in the same direction

Case 2: Moving Sideways

All wheels are turned 900 or 2700 and then all wheels rotate at the same speed



page 21/54

Figure 11: Cases of Wheels orientation towards direction of moving

This kind of wheels are ideal for assuring that the trolley will have a good traction, without loss by friction and without slipping so that a very good odometry could be applied. The disadvantages to the selection of such type of wheels are that their mechanical construction is difficult to achieve, while it is also difficult to program them. Upon appropriate programming the user could have full control in the movements in all the directions and including rotations.

Case 3: Moving Around a Bend

All wheels are turned in the appropriate angle and are moved with the same speed

Case 5: Rotation around the Central Point of One Axle

All wheels are turned to the required direction and only the front or the back wheels are turning with the same speed. The remain wheels are breaking of one axle only rotate in

opposite directions but at equal speeds

Case 4: Rotation

Two or all the wheels are turned constantly and moved with the same speed



page 22/54

The first concept design towards this rotator wheels is displayed in Figure 12. We should mention that this design as well as the design of the overall trolley have been performed and analysed in order to be very close to the manufacturing phase of the trolley and its sub-systems.

Figure 12: Rotary Wheels Mechanism – Concept Design

In this concept design a yaw and rotation housing with gearboxes have been included in the cylindrical upper component of this design. In every wheel a single motor with an incremental encoder has been integrated. The first motor (in the upper part) has been utilized for turning the wheel in the appropriate direction, while the second motor (in the lower part) has been utilized for giving the necessary rotation to the wheel for achieving the movement.

3.2.2 Specifications The vehicle’s motion will be based on the combined control of four two-axis wheels. Each wheel will support a forward movement and a self-rotation to accomplish successfully the most complicated and sophisticated driving tasks. The motors should be able to move the trolley with at least 2m/s under full load, even if driving on a slight slope-up. Additionally the motors should be able to accelerate the robot in a reasonable time. Based on the previous



page 23/54

mentioned needs the motors for the drive of each shaft will be dc motor type and rated about 100 Watts, and encoders of 1000 ppr will provide the desirable resolution in the position-speed control of the wheel.

In our application the need of braking is of paramount importance. The braking action will be especially necessary upon an identification of an obstacle or whenever a safety condition has been raised, that needs to slow down or emergency stop the trolley. The brakes should be able to block the wheels of the robots at once, even if fully loaded. For achieving the braking action an H-Bridge configuration will be needed. For the construction of this braking mechanism specific advanced and integrated electronic control boards will be developed.

The control action of each separate motor will be based on the classical Proportional Integral Derivative (PID) control algorithm. The algorithm will be running on an embedded computer board and will be connected with the motors via the CAN bus. In Figure 13 we present the under design embedded electronic board (designed by ZENON S.A.) that will be utilized for controlling (reading the encoders and braking) the motors. The size of the board has been intentionally kept small to fit into the small housing of the rotary wheels.

Figure 13: Integrated Controller Board

3.3 Energy supply

The trolley’s electric and electronics systems (including sensors, PCs, controllers and actuators) will be powered through two or three (if necessary) heavy duty 12Volts batteries in serial connection to provide 24Volts. These batteries are presented in figure 14 and must be sealed and rated for high currents, long durability and lifetime, in order to comply with the power requirements of the whole system. Furthermore, the applied voltage from the batteries will be separately handled by each subsystem to convert the input voltage to the specified voltage it needs. The estimated energy supply will be calculated to provide a sufficient operation of the trolley, in full operation, for minimum 1 hour.



page 24/54

Figure 14: Energy Supply Unit

The accumulators will be charged with a standard charger for this type of accumulator. There is no special charging concept as e.g. an automatic drive of the trolley to a charging station. There will be manual connection outside of the vehicle from charger and accumulator. The energy concept does not foresee an optimisation of the energy efficiency of components because of the research character of CommRob, addressing other topics. Having special concepts for power efficiency would mean additional difficulty which would not be helpful for realising them. The exact type of accumulator will be specified in the next stage of the project.

3.4 Warning and Safety devices The trolley will drive in several modes autonomously and therefore needs special warning and safety devices. These devices will follow the AGV guideline 4451/6 from VDI [2], where it is mentioned that the securing of the AGV includes the protection of personnel and the protection of the plant components and of material assets in the environment of the vehicle. The safety and warning devices will be used in all modes to ensure safety when moving the trolley.

In this case the need of having sensory data is mandatory so that hazards can be detected and avoided automatically. This needs to be ensured by the devices that will be presented in the following subsections.

3.4.1 Person detection system It is the task of this system to recognize persons in its way and to generate a reliable signal for braking the trolley. There are two possible methods for doing so, which can be used either single or combined.

3.4.1.1 Laser scanner The Laser scanner uses an infrared light beam which is deflected by a rotating mirror and so establishes a measuring area in its front. Objects in its way will reflect this light and send it back to an integrated receiver. In the span of time between sending the beam and measuring the reflection the laser scanner calculates the distance to the object. Moreover, the laser scanner can detect the direction of the object, reflecting the beam.

Detecting an object within a definable protected area leads then to the sending of a signal which can be used for stopping the trolley. A second area can be defined as warning area. Detecting an object within this area leads to the sending of a different signal which can be



page 25/54

used e.g. for slowing down the trolley. Figure 15 shows a vehicle with mounted laser scanner and shows the security and warning area in driving direction.

Figure 15: Automated guided vehicle with laser scanner

In CommRob, the laser scanner S300 from Sick will be used. For detailed specification of this sensor please see the chapter 5.7.2.

There is currently planned to mount two laser scanners in the front and back of the vehicle which are placed in opposite corners. With the scanners having a measurement angle of 270°, it will be possible to have an overall field of view of 360° around the trolley. This secures the whole area around the trolley and allows to drive in all directions.

3.4.1.2 Bumper The second possibility for person detection is the usage of bumpers. Bumpers are made of a foam core and have a safety element within, which is pressure-sensitive. So the bumper generates an output signal by having direct contact with an object. This signal can then be further processed to initiate an emergency stop.

The bumpers used on the trolley will ensure that all driving directions are completely secured, as they are mounted around the trolley.

A possible provider of bumpers is ASO Safety Solutions.

3.4.1.3 Multi-camera based system In chapter 5.7.1 is a multi-camera based system presented for obstacle detection. The possibilities in using this device for reliable safety signals will be evaluated at LAAS.

3.4.2 Braking system The braking system needs to be able to stop the vehicle dead before any solid part of the trolley or of its load touches the person.

Moreover there will be an integrated self-acting brake system in the motors, so that the trolley can not be moved by loss of energy supply (please see chapter 3.2.1).

Protected area (dark grey) and warning area (light grey)



page 26/54

3.4.3 Warning devices The trolley will be equipped with optical signals for showing vehicle traction standby and drive as it is displayed in Figures 5 and 6 at the lower part of the trolley’s chassis

There will also be optical signals for showing changes in the driving direction of the trolley while the exact type of optical devices will be specified in the next stage of the project.

3.4.4 Emergency stop There will be mounted an emergency stop which is accessible anytime. Hitting this stop button will stop the trolley dead as defined by the braking system. The exact type of emergency stop will be specified in the next stage of the project.

3.5 Operating elements

3.5.1 Handle User Interface For the design of the Handle User Interface (HUI), four cases could be incorporated at the same time (depending of the user and its needs). These four options are: a) a switch user interface, b) a voice user interface, c) a handles user interface, and d) the user feedback. The final concept of the handle user interface is presented in Figure 16. Moreover in this figure the microphone array concept is also presented. This array will be operated as it is analyzed in Section 5.1 and in the deliverables D4.1, D4.2.1 and D4.3.1.

Figure 16: Haptic handle concept drawing



page 27/54

With respect to the walking aid capabilities of the CommRob trolley, the user interface could be designed to enable two driving postures. The first is the normal posture where the user is handling the device from the two handlebars as he/she would do with a traditional zimmer frame.

The second driving posture- the relaxed posture- enables the user to put all his/her weight on the device. The position and shape prevent the user from sliding from the device, as in the case that it is displayed in Figure 17. This picture has been taken from a walking aid EU funded research project that was called PAMAID and is displaying the mentioned concepts.

Figure 17: Modes of driving postures

Except from the touch screen panel, the device may alternative be driven by the use of a set of switches located at the top of the user interface. All the switches required to drive the device are in the view of the user and are easily located to positions for those users who are unable to see or have limited sight abilities.

For the normal postures the right and the left turn switches where placed in the inner part of the user interface. The placement of the switches enables the user to apply on the handles any force required in order to support his or her weight, while h/she is still able to operate the switch. The user interface is displayed in Figure 15.

In the following subsections, the four handle user interfaces will be presented.



page 28/54

3.5.1.1 Switch User Interface Additional to the touch screen interface the user’s input is acquired by a set of switches. Their names and functionality could be like the following:

• On/Off Switch: Starts and stops the operation of the device. It is a two position switch equipped with an associated light that indicates the on, off states, which are also indicated by the position of the switch.

• Emergency Stop: Pause the motion of the device. It is a two-position switch, where one position is locked.

3.5.1.2 Voice User Interface For the case where the CommRob will work in the Walking Aiding mode the user maybe possible not to be able to perform control of all the HMI’s control buttons provided. For this case the voice user interface will be of great usage for the disabled people. In this interface the voice user interface will incorporate vocal commands that have identical functionality to the switch user interface.

3.5.1.3 Handles User Interface The handles user interface differs from the switch user interface in the way the left and the right instruction are given. All switches of the switch user interface are remaining except the left and the right switches. The switches of the handles user interface have identical functionality to the switch user interface. The left and the right switches are replaced by handles. In manual mode, a left (right) turn of the handles produces a linear mapping between the handles turn angle and the steering angle.

3.5.1.4 User Feedback The user feedback addresses the acts of the user in order to interactively communicate with the trolley. The most common ways is via voice messages (generated from the trolley) or tune messages. In addition to those another type of user feedback will be very useful – the user input confirmation. The main usage of this functionality is the confirmation of the user that the given command (through any kind of user interface) has been accepted or rejected from the trolley as appropriate.

3.5.2 Human IN/OUT The main Human In/Out component (can be also mentioned as an integrated Graphical User Interface – GUI) will be the main component that the user of the trolley will have the ability to exchange data and commands to the trolley as also get informed about the current situation of the trolley, the mode of operation and of any other valuable information. This component will base its operation in a touch screen panel as the one that it is displayed in Figure 18. These panels based on the ability to detect the location of the touches within the display area will allow their usage also as an input device that will remove the keyboard and the mouse from acting as primary input devices for interacting with the display content.



page 29/54

Figure 18: The touch screen panel – Human In\Out interface

The touch screen panel will be a resistive or a capacitive touch panel and will be of minimum a diagonal of 14 inches. Regarding the size of the panel a compromising study should be made in order to select among having a bigger panel for the user or having a smaller panel that will increase the field view of the user. Moreover this touch screen panel will be put on a passive mechanical pan-tilt platform for allowing the customization of the user’s field of view to the screen. This pan-tilt unit is displayed in the previous Figure (at the bottom) and its calibration will be made by the user.



page 30/54

4. Trolley computer based systems The following paragraphs specify the requirements of each computer based subsystem as well as the Motion Control system and the wireless communication system. The specification in this document shows the minimum requirements and the exact type of PCs will be specified in the next stage of the project and especially during the phases of the software module developments.

4.1 PC for communication Type of PC: PC

Intel-Processor: Dual core 3GHz

Operating system: Linux

RAM: 1GB

Interfaces: Firewire, USB 2.0, Ethernet

Hard-disk memory: 120 GByte

Further SW: TUW communication software

Dimensions: 520mm x 210mm x 510mm (H x W x L)

4.2 PC for planning and interaction Type of PC: PC



RAM: 1GB



Further SW: LAAS Speech recognition software




page 31/54

4.3 PC for Execution Type of PC: PC



RAM: 1GB



Further SW: FZI/LAAS execution software


4.4 Motion Control System As mentioned in previous paragraphs, the motion of the trolley is based on the combined action of the four wheels. Therefore the major task for the control algorithm of the motion is to translate the desired path of the trolley to the appropriate commands in every wheel, in terms of rotation and travel actions. This can be realized by a system (Figure 19) which consists of a master – multiple slaves architecture. The master in this occasion is the PC which issues several commands to the supervisory controller related with the overall speed, direction and position of the trolley. Then, the supervisory controller translates these general commands to control signals for every wheel. This complicated task requires from the supervisory controller to solve a mathematical equation system that will provide the exact drive signals for every motor in each wheel, so that the combined action of them will result in the desired manoeuvre of the trolley. In the proposed topology, the motor controllers will establish a reliable connection with the supervisory controller through a robust communication bus such as CANbus or RS485. Furthermore, the motor controllers will be individually addressed in the bus in order to be separately commanded by the supervisory master controller.



page 32/54

1st WHEEL 2nd WHEEL

M

M

MOTORCONTROLLERS

MOTORSCOMMAND

ENCODERSFEEDBACK M

M

MOTORCONTROLLERS

MOTORSCOMMAND

ENCODERSFEEDBACK

BACK WHEELS 1st WHEEL 2nd WHEEL

M

M

MOTORCONTROLLERS

MOTORSCOMMAND

ENCODERSFEEDBACK M

M

MOTORCONTROLLERS

MOTORSCOMMAND

ENCODERSFEEDBACK

FRONT WHEELS

PC SUPERVISORY CONTROLLER

VEHICLE POSITION – SPEED

SETPOINT

COMMUNICATION BUS

Figure 19: Motion Control System

Figure 20: Motion Control System physical mapping

The 8 Controllers presented in Figure 20 are the in-house under design DC brushed motor controllers while the 8 DC motors are going to be DC Brushed motors with a incremental encoder and an appropriate gearbox as specified in paragraph 3.2.2 specifications. This



page 33/54

supervisory controller is one of the in-house under design motor controllers that will be programmed to provide a bridge between the communication protocols (CAN and Serial RS232). It will take data from the PC in terms of rotational position and speed value and translate them to the other controller via an appropriate protocol on the CAN bus.

4.5 Wireless Communication For supporting the wireless communication among the different processing units of the trolley an IEEE 802.11b WLAN switch will be utilized. This communication channel will be also utilized for providing access for the Maintenance Station and for robot-robot communication. The communication channel will support both peer to peer and broadcasting capabilities while will achieve a severe reduction in the necessary cables needed.



page 34/54

5. Logical sensors

5.1 Audio In

5.1.1.1 Audio device The exploitation of sound cues will be based on an integrated auditory sensor developed at LAAS-CNRS. From the spatial sampling of the acoustic wave fields by an array of 8 microphones (figure 21), an acoustic energy map will be computed, enabling the localization of the sound sources in the environment. This “low-level” auditory cue will serve as a basis to (1) the focalization of the sensor (through its “electronic polarization”) towards specific spatial areas in order to extract the signal emitted by the robot’s user, and (2) the fusion of vision and audition for tracking.

Geometry Linear array of eight microphones, at inter space λ3kHz/2=5.6cm Array total length < 40cm

Microphones characteristics (in the research prototype)

¼” GRAS 40PQ Sensitivity@250Hz = 8mV/Pa Frequency response & Phase match: ±1dB & ±1° over [100Hz;3kHz] Output voltage for a 60dB conversation (~20mPa) ≈ 0.16mV

Acquisition chain (reusable)

Fully programmable 8-channels acquisition chain developed at LAAS (High-pass filter, Programmable Gain Amplifier, Programmable Anti-aliasing filter, ADC@Programmable sampling rate) Acquisition@15kHz

Processing unit (reusable)

Virtex 4 SX35 Xilinx FPGA evaluation board Localization@15Hz

Communication (reusable) USB1.1 – Driver library developed at LAAS

Power Supply

Any compatible Constant Current Power Supply Can be connected to a 24V DC battery Current consumption < 2A Power consumption: ~ 20W

Table 1: Characteristics of the auditory sensor.

Figure 21: The array of 8 microphones developed at LAAS-CNRS.



page 35/54

5.2 Audio Out The audio out capabilities will be fulfilled by a set of stereophonic sound speakers. These speakers are very famous in the market and no further exact specification should be given.

5.3 RFID

5.3.1 RFID for user identification RFID technology is currently investigated, as on-board readers can enable the trolley to detect in its vicinity Users wearing RFID tags. LAAS-CNRS has been evaluating the RFID setup described in table 2 while the antennas fixed on a robotic platform are shown on figure 22 The size of the antennas and tags can be divided by 3 with an equivalent system working at 2.4GHz (see the URL www.synometrix.com).

Reader 870 MHz frequency, multi-protocol, programmable emitting power

Tag Passive, ISO 18000-6B compliant, 196B free space, almost unlimited rewriting capacity

Antennas Eight 24cmx24cm flat antennas with 60°x75° each Power consumption

1A@15V (TX/RX mode), 260mA@15V (idle mode)

Type CAENRFID A941

Table 2: Example of RFID device from LAAS-CNRS

Figure 22: RFID setup mounted on a platform at LAAS-CNRS



page 36/54

Such system, by addressing sequentially all the antennas, will detect the User all around the robot at H/R distance in the range of 1 to 5 meters. We are using eight antennas, placed on a regular octagon (each edge is around 30cm). We have to make some hardware adjustments to control the 8 antennas, because our reader is able to control only 4 antennas. We placed switches controlled by an unused pin of the RS232 port connected to the reader. Preliminary evaluations at LAAS highlight that, thanks to this strategy, the major detection range consists of a pan arc with an opening angle of 40° in the direction of each antenna. Evaluations are also pursued to provide distance information based on the time required to receive the tag response. Another RFID setup for robot navigation is also evaluated by the FZI team.

5.3.2 RFID for navigation

5.3.2.1 General principle The system of navigation by RFID consists of a RFID-Reader which is placed beneath the robot and RFID-tags which are placed on the floor, so that the trolley recognises them when driving over. The placement of the tags will be in doorways or across corridors and divides a big area into several sub-areas according to the semantic navigation.

The principle is illustrated in the following figure:

Figure 23: Scheme for RFID based navigation in CommRob

Thereby it is also possible to determine the driving direction by using double-tags and reading both of them sequentially.



page 37/54

5.3.2.2 Specification In CommRob the RFID-Reader ID ISC PR101 will be used which has the following specification:

Reader FEIG Proximity Reader ID ISC.PR101-A/-USB

Housing Plastic ABS Colour Papyrus White RAL 9018 Dimensions (WxLxH) 85 x 145 x 31 mm (3.35 x 4.72 x 1.77 inch) Protection class IP 30 Weight 200 g (0.44 lb) Supply voltage - variant –A (RS232/RS485) - variant -USB

typical 12 V DC max. 12 - 24 V DC +/- 15% 5 V DC (via USB)

Current draw max. 0.5 A Power consumption - variant –A (RS232/RS485) - variant -USB

max. 5 VA max. 2,5 VA

Operating frequency 13.56 MHz Transmitting power 0,5 W +/- 2dB Antenna integrated Reading distance max. 18 cm Interfaces RS232 / RS485 (configurable)

USB (12 Mbit) Optical indicator 1 LED (multicolour; red/green) Protocol Modes FEIG ISO HOST & Scan Mode Supported transponders - ISO15693, ISO18000-3-Mode1

(EM HF ISO chips, Fujitsu HF ISO chips, KSW Sensor chips, Infineon my-d, NXP I-Code, STM LRI ISO chips, TI Tag-it) - NXP I-Code1, I-Code UID, I-Code EPC

Address setting for interface - Variant –A (RS232/RS485) - variant -USB

Software (up to 254 addresses) Device ID of the reader

Temperature range - operation - storage

-25°C to 60°C (-13°F to 140°F) -25°C to 70°C (-13°F to 185°F)

Humidity 5 - 95% (non condensing)

Table 3: Characteristics of the RFID-Reader

Sensor evaluation showed that there is a coupling area of the RFID-reader at 10 cm distance which has approximately 15 cm in diameter and another coupling area at 2 cm distance with approximately 30 cm in diameter. This is illustrated in figure 24 and figure 25.



page 38/54

Figure 24: RFID measurement coil

Figure 25: RFID measurement area

The bigger measuring area at 2 cm distance reasons the mounting of the RFID-reader at 2cm above the ground.

When using a ‘two tag barrier’ at least one tag is detected with 100% reliability up to a speed of 2m/s as shown in figure 26.

Figure 26: Reading performance of a “double” RFID tag



page 39/54

The placement of the tags has to be with up to 10 cm between them without loosing reliability. If both tags are read the driving direction can be determined. A possible set-up is shown in figure 27.

Figure 27: Distance for placement of RFID tags

5.4 Wireless

5.4.1.1 Introduction Traditionally, devices in a radiolocation system measure some characteristic of the RF signal to or from fixed access points, such as Received Signal Strength (RSS), time of arrival, or angle of arrival. These measurements are subsequently used to estimate the location of individual devices by solving systems of equations based on geometric principles. Compared to time of arrival and angle of arrival, RSS is extremely attractive for practical realization in Wireless Sensor Networks (WSNs) because of the simplicity of its implementation. Despite its simplicity, however, practical application of RSS to radiolocation has been limited due to the intrinsically low ranging accuracy. Recently a new class of algorithm, referred to as relative location, has emerged, achieving improved accuracy of location estimations using RSS measurements. A Location Engine is used to estimate the position of nodes in an ad-hoc wireless network. Reference nodes are placed with known coordinates, typically because they are part of an installed infrastructure. Other nodes are blind nodes, whose coordinates need to be estimated. These blind nodes are often mobile and attached to assets that need to be tracked. The self localization system based on ZigBee will deliver the position of the robot and the position of the user while there will be a 'global' coordinate system in the super market.

5.4.1.2 Proposed Self Locating System For self locating of the robot a wireless sensor network will be used, where nodes will have a location engine to estimate the location. Sensors will be placed on some of the shelves and on each robot. Since the position of some of the sensors placed on the shelves is known the

10cm15cm

100cm



page 40/54

position of the robots can be determined. The information of the robot’s position is used for navigating the robot through the store and to certain positions. Navigation is supported by additional sensors on the robot not described in more detail in this document.

The user itself will also get a small wireless communication device, which locates itself and transfers the user’s position to the assigned robot. This position information is used for following the user and finding the user in the store. The handheld device has also a button for calling the robot. In addition the sensor network can be used for transmitting information between the shelves and the robot. Since this information sharing does not require high bandwidth low to medium bit-rate transceivers are sufficient

Figure 28: Concept Application of the WSN for Indoor Localization

As it can be seen in Figure 28 the overall system consists of reference nodes placed on the shelves on at least one node in each robot and a node like a button for the user to be put into the pocket. The nodes can be of the same type allowing for a fully homogeneous hardware platform. Shelves are typically set up in long rows separated only a few meters (>3m) from each other while the nodes shall be placed on top of the shelves in a distance of about 5 meter. This gives a rectangular grid of 3x5m. In a supermarket of 1500m2 this gives about 150 nodes while in CommRob a few nodes will be utilized for proving the concept of the localization based on WSN in a small scale demo (reduced geographical area).

5.4.1.3 The sensor node State-of-the art shows that different technologies can be used for indoor node localization when the position of some reference stations is well known. Technologies based on Ultra Wideband (UWB) have not been taken into account due to the short range and non-maturity of the technology.



page 41/54

Three different Chip solutions are proposed. Both shall be investigated in more detail, to find out which one best satisfies the requirements and proves to be more robust.

1. CC2431 ZigBee based SoC with location engine from Texas Instruments1 2. NanoLoc Transceiver from the Company Nanotron2 3. The Cricket mote from the Company Crossbow3

In the following paragraph the three solutions are described in more detail:

5.4.1.3.1 CC2421 from Texas Instruments: The CC2431 with Location Engine (Figure 29) is a true System-on-Chip (SoC) solution for ZigBee/IEEE 802.15.4 wireless sensor networking. The CC2431 combines the known CC2420 RF transceiver with an enhanced 8051 MCU, 8 kB of RAM, 128 kB Flash memory and many low power features. The Chip was developed by the former Company Chipcon, which has recently been acquired by Texas Instruments.

Figure 29: The Chipcon mote

The Location Engine implements a distributed computation algorithm that uses received signal strength indicator (RSSI) values from known reference nodes, such as mobile neighbour nodes with the same Location Engine or fixed infrastructure nodes. Performing position calculations at the node level reduces network traffic and communications delays otherwise present in a centralized computation approach.

The Location Engine has the following main features:

1 The Texas Instrument Mote: http://www.ti.com/corp/docs/landing/cc2431/index.htm

2The Nanotrone Mote: http://www.nanotron.com/EN/NE_releases_2006-10-09-DE.php

3 The Cricket Mote: http://www.xbow.com/Products/productdetails.aspx?sid=176



page 42/54

• A blind node can use from three to eight reference nodes for the location estimation algorithm

• Location estimate with resolution of 0.5 meters

• Time to estimate node location less than 40 µs

• Location range 64 x 64 meters

• Runs location estimation with minimum CPU usage

• Location error can be less than 3 meters, depending on factors described below

It is stated in the manuals that in order to achieve the best possible accuracy one should use antennas that have near-isotropic radiation characteristics. The location error depends on signal environment, deployment pattern of reference nodes and the density of reference nodes in a given area. In general, having more reference nodes available improves the accuracy of the location estimation.

Advantages: It is the market's lowest cost solution for positioning in low power wireless sensor networking. It is the only ZigBee compliant solution available with integrated location engine.

Disadvantages: Comparable to Nanotron lower accuracy Multiple reference stations needed for lowering the error. 5.4.1.3.2 NanoLoc Transceiver from the Company Nanotron The nanoLOC TRX Transceiver (Figure 30) is a highly integrated chip utilizing Nanotron’s wireless communication

Technology CSS (Chirp Spread Spectrum). With its Chirp ranging capability, nanoLOC can measure the link distance between two nodes. Thus, nanoLOC supports location-aware applications including Location Based Services (LBS), enhanced RFID, as well as asset tracking (2D/3D RTLS).

Figure 30: The nanoLOC TRX Transceiver



page 43/54

Chip features nanoLOC supports an adjustable centre frequency with 3 non-overlapping frequency channels. This provides support for multiple physically independent networks and improved coexistence with existing 2.4GHz systems. Data rates are selectable from 2 Mbps to 125 kbps. Due to the chip’s unique chirp pulse, adjustment of the antenna is not critical.

• Logistics: Asset tracking / RTLS / Active RFID

• 2D / 3D Real Time Location Systems (RTLS)

• Built-in precise ranging: 2 m indoors, 1 m outdoors

• Operates worldwide: 2.45 GHz ISM band

• Data rates: 2 Mbps to 125 kbps

• Modulation technique: Chirp Spread Spectrum

SDS-TWR Ranging Algorithm Nanotron's SDS-TWR algorithm (Symmetrical Double- Sided Two-Way Ranging) allows better accuracy than the CC2431 based on datasheet comparison even with the use of low cost crystals for the oscillators. High frequency clocks are not required to achieve a resolution of less than a meter.

Advantages: Higher accuracy and robustness More flexibility in choosing an appropriate Microcontroller Disadvantages: Much higher chip costs than the CC2431 Need for an external Microcontroller 5.4.1.3.3 Mica2 Cricket Mote from Crossbow The Mica2 Cricket Mote, is a location aware version of the popular Mica2 low-power Processor/Radio module. This mote includes all the standard software from Crossbow and an ultrasound transmitter and receiver. This device is using the combination of RF and Ultrasound technologies to establish differential time of arrival and hence linear range estimates.

These motes can be configured as both listeners and beacons. The beacons are placed throughout a building or facility and transmit concurrent RF and ultrasound pulses. Listeners are attached to mobile devices and listen for RF signals. Upon receipt of the beacon RF signals, the listener then listens for the corresponding ultrasonic pulse. When this pulse arrives, the listener obtains a distance estimate for the corresponding beacon by taking advantage of the difference in propagation speed between RF and ultrasound. The listener runs algorithms that correlate the RF and ultrasound samples to pick the best correlation.



page 44/54

Advantages: Even in the presence of several competing beacon transmissions, the mote rapidly achieves impressive precision and accuracy. Disadvantages: High costs

Figure 31: The MICA2 Cricket Location Mote

5.4.1.4 The sensor node placed on the shelves The most accurate way of giving the node its reference location is storing measured (X,Y) coordinates from a given (0,0) location within each node and placing the node accordingly. The disadvantage is that each node has to be programmed separately and nodes cannot just be randomly placed. A better solution is to store a list with the entire floor plan of (X,Y) coordinates of every node in each node. So every node has the same program inside. After the network setup an algorithm should be started where each node identifies its entry in the list. The algorithm can start from the node closest to the (0,0) position (needs to be triggered by an installation person at the end of the physical installation) where each nodes claims to be owner of an entry in the list based on relative distance measurements. Such an algorithm is not elaborated in detail yet but is part of the future task list. Note that only a two dimensional plane will be taken into account.

The reference nodes periodically send a beacon with its position information and transmit power as a broadcast message. The reference nodes also build up a network where they try to synchronize each other so that every reference node sends its position information in a different slot. A mobile node tries to identify all messages of reference nodes it gets and calculates its position with an algorithm which is dependent of the technology used. In our case the mote platforms are coming from: a) Nanotron, b) Texas Instruments, and c) Crossbow.

5.4.1.5 The sensor node included in the robot Each robot will be equipped with a sensor node. This node locates itself and gets information of the user’s position from the assigned user’s positioning node. This node is powered by the power supply of the robot. The sensor node in the robot can be the same as the sensor nodes on the shelves. The position information of the user is transmitted to the User Perception Module. The same applies to the robot’s position. The command of a user calling its robot is transmitted via the User Perception Module to the communication unit.



page 45/54

5.4.1.6 The sensor node on the user The sensor node carried by the user Each user will get a device for localizing the user. This sensor node will transmit the position of the user to its dedicated robot's sensor node using the radio transmission capability of the sensor node. This information will be used to support the user identification.



page 46/54

5.5 Visual Features

5.5.1 Stereo camera for the vision on the user The stereo system will be placed on a pan and tilt platform which is mounted on top of a mast, so that occlusions will be limited. The focal lengths and the stereo baseline will be revisited with respect to the limitation of the pan rotation (i.e. only 270° on the Directed Perception platform used on our robot), so that the robot could track the user whatever his position around the robot.

Current investigations at LAAS rely on the following Pan Tilt platform (Table 4).

Imaging sensor Mono 2/3” CCD camera, mosaic of bayer filters

Pixel array resolution 640x480 View field per camera 88° x 72° Power consumption per camera Less than 2W

Interface IEEE1394b Baseline between the two cameras 20cm to 30 cm

Camera type Flea2, Point Grey Research: www.ptgrey.com Pan unit type Directed Perception

Table 4: Characteristics of the Pan Tilt platform

Such a system (which is assumed off-line calibrated) and dense stereo-correlation algorithms will output disparity map corresponding measure H/R distances in the range [1 - 5] meters.

During the project, there will be studied possible alternatives to this approach (see [5]). One alternative is using a Time-Of-Flight Optical camera with the two cameras rigidly mounted on the pan and tilt platform.

Advantages: Acquisition of dense 3D images at video frame; more robust to illumination changes and lack of texture in the scene; fusion with photometric information given by the camera

Disadvantages: High cost of the TOF sensors



page 47/54

Investigations at LAAS are performed with the range imaging camera as described in table 5:

Pixel array resolution 176x144 pixels Field of view 47.5° x 39.6° Interface USB2.0 Illumination power 1W at 850 nm Power 12V, 1A Output data x, y, z coordinates Time-of-light camera type

Swiss ranger SR3000, Mesa Imaging, see for more details the URL www.mesa-imaging.ch

Table 5: Characteristics of the range imaging camera used at LAAS-CNRS

Figure 32 shows two flea2 cameras in stereoscopic configuration, together with a Swiss ranger camera fixed in the middle of the baseline.

Figure 32: Stereo and range imaging camera setup at LAAS-CNRS

5.5.2 Camera for the robot localization This camera will be static and mounted on a mast in such a position that it does not prevent the stereo system from remaining oriented towards the user, whatever his position around the robot. The camera could be a Point Grey Flea2 (see [5]).

A lens with a short focal length (i.e. 3.6mm) will be chosen with respect to the usual width of a corridor in a commercial centre. This will ensure that the robot can detect visual landmarks on both sides.



page 48/54

During the project, there will be studied possible alternatives to this approach e.g a smaller and cheaper colour camera, like the ones mounted on mobile phones or on lap tops:

Advantages: Price, compactness

Disadvantages: No flexibility on the focal length, image quality

A further alternative is the usage of a PTZ camera:

Advantages: More flexibility to select the view filed on the right or on the left of a corridor, depending on the product to be reached, better resolution.

Disadvantages: More expensive, more complex.

5.5.3 Camera for obstacle detection The obstacle detection and local modelling will be done using a robot-centred occupancy grid. There will be used a belt of micro-cameras all around the robot: at least 4 but probably more (i.e. 8) cameras with fish-eyes for a quadrangular robot. They will be equipped with lenses which have a focal that is longer than the focal of the lenses used for the camera dedicated to the robot localization.

Every camera detects the free ground around the robot. The acquisitions are synchronized and all free areas are fused when updating the occupancy grid. This will be done by a dedicated hardware (providing a logical sensor) which processes all images without consuming the PC computation resources.

Cameras could be provided by STMicroelectronics or other specialists on micro-cameras for security applications.

A more detailed description on the function of the system can be found in [5].



page 49/54

5.6 Haptics The haptic device consists of a 3D force torque sensor and two sensors for the detection of contact to the user. The 3D force torque sensor measures two orthogonal forces in the horizontal plane of the trolley. It also measures the torque around the axis normal to the horizontal plane. The measurement will be carried out by the use of strain gauges. With these three values it is possible to extract all information necessary to steer the trolley in an omnidirectional manner. The displacement as well as the change of orientation of the trolley by the user can be sensed. With those values for example a zero force control is possible. Additional to the 3D force torque sensor two sensors for detection of contact are implemented at the trolley handle. These two sensors are (in that stage of the project) simple on/off switches which detect the contact of each hand of the user to the trolley.

5.6.1 Mechanical design of the haptic handle A first concept of the mechanical design of the haptic handle has been described in detail in D4.2.1 [3]. The current setup of the haptic handle focuses on the elementary requirements which have been described above. This setup can be used as a haptic handle for the robot trolley as well as a sensor of a common trolley to gather sensor information for example for the user centred prototyping evaluation. The setup of the haptic handle mounted on a common trolley is shown in the left picture of Figure 33 while in Section 3.5.1 an alternative investigation is presented. Both concepts are of the same operation regarding the strain gauges beams.

Figure 33 (left) CAD drawing of the haptic handle (right) Placement of the strain gages on the different strain gauges beams (B1-B4)

The main part which hosts the force and torque sensor consists of two planes, which are connected by four strain gauges beams. One plane is directly connected to the trolley. The second plane carries the handle, which will be used by the user to steer the trolley. Due to this setup the flow of force must go from the handle through the second plane, the strain gauges beams and finally the first plane to the trolley.

The strain gauges for the detection of the forces and the torques are placed on three of the four strain gauges beams. Figure 33 on the right illustrates the positions of the different strain gauges and the force direction they sense. The beams B1 and B2 are equipped with strain gauges to sense the forces in the x-direction. The beams B1 and B3 are equipped with strain



page 50/54

gauges to sense the forces in the y-direction. The number of strain gauges is sufficient because in general three forces (with independent force directions) will be enough to calculate the forces and the torque of the haptic handle.

5.6.2 Electronic components of the haptic handle The electronic components of the haptic handle are responsible for the measurement of the strain gauges, the measurement of the human contact, the prepossessing of the gathered sensor information as well as the communication with the overall system. Figure 34 illustrates the used components and their connections.

Figure 34: Schematic diagram on the electronic components of the haptic handle

The major part of the components is the so call UCoM, which is responsible for the gathering of the sensor information as well for the prepossessing of the sensor information. This board has been developed at the FZI as a modular hardware component for example for lower level control tasks (motor control) as well as sensor prepossession. A detailed description on the UCoM can be found in [4]. The UCoM communicates with the overall system over CAN-BUS. The UCoM is directly connected to the sensors for the contact of the human, which are on-off switches one for each hand.

For the measurement of the strain gauges the specialised component “PS021 - Digital Amplifier for Strain Gages” from acam-messelectronic gmbh (http://www.acam.de) is used. The main advantage of this component is that the integrated circuits are realized without any analogue component but an approved Time-to-Digital Converter (TDC) technology. This technology has a lot of advantages like there is no need for a full-bridge, 2 resistors (half-bridge) are sufficient or no reference voltage is needed. The PS021is connected to the UCoM over SPI-BUS.

5.6.3 Prototype and first tests of the haptic handle The parts for the haptic handle have been manufactured and the first prototype has been assembled. The sensor for the contact of the user will be embedded on the haptic handle after the detailed tests of the strain gauges setup. The prototype of the haptic handle is shown in figure 35.



page 51/54

Figure 35: (left) Prototype of the haptic handle (right) Position of the strain gauges with measurement electronics

5.7 Distance

5.7.1 Device for security In the field of automated guided vehicles, specific rules for safety and warning devices must be respected as described in chapter 3.4.

For the person detection system, the laser scanner S300 from Sick will be introduced (please see figure 36 and table 6 for specification).

If using the laser scanner there will be need of two laser scanners with a 270deg field of view. These should be located in opposite corners of the trolley. This will give the trolley a 360deg field of view.

Figure 36: Laser scanner Sick S300 Professional



page 52/54

Laserscanner Sick S300 Professional Ambient operating temperature from ... to: -10 ... 50 °C Dimensions (W x H x D): 102 x 152 x 105 mm Weight: 1.2 kg Scan angle: 270 ° Protective field range: 2 m Maximum warning field range: 8 m Number of field sets: 4 Response time: 80 ms Resolution: 30 mm, 40 mm, 50 mm, 70 mm, selectable Distance measuring range: 30 m Number of mulitple samplings: configurable via CDS 2 ... 16

Electrical connection: Plug-in connection housing with screw, Screw-type terminals

Supply voltage: 16.8 V DC, 24 V DC, 30 V DC Number of EDM inputs: 1 Number of reset-/restart inputs: 1 Number of static or dynamic control inputs: 2 Number of stand-by inputs: 1 Number of safe outputs: 2 , Safety outputs (OSSD) Number of warning field outputs: 1 Number of application diagnostic outputs: 1 Maximum output current: 250 mA Configuration and diagnostics interface: RS-232 Data interface: RS-422

Table 6: Characteristics of the laser scanner Sick S300

5.7.2 Device for obstacle detection The obstacle detection and local modelling will be done using a robot-centred occupancy grid. The logical sensor for this task can be a SICK Laser Range Finder mounted forward on the trolley. Beneath the advantages of this system, there exist several disadvantages. As it might not be possible to detect obstacles all around the robot, usage of one single sensor will not be sufficient. As the price for the sensor is quite high, the usage of several of them will be expensive. Moreover, using laser technology could be considered as dangerous in a public environment like a store.

A solution for this is evaluated on a robotics platform at LAAS-CNRS using micro-cameras around the robot. This solution is described in chapter 5.5.3.



page 53/54

6. Conclusions The deliverable shows a first design specification which describes the CommRob trolleys in detail. This deliverable is a live document as during the evolution of the project this will be updated and upgraded in order to include all the technological approaches that have been inserted or developed to fulfil the requirements of the CommRob trolley. At the end of the project, the updated document will contain the specification of the thoroughly tested trolley system. In this deliverable the first design concept of the trolley and of all the needed sub-modules have been detailed described. This description is in full compliance with the efforts and approaches reached until now and with the conclusions drawn from the other WPs.



page 54/54

7. References [1] CommRob Deliverable D2.1. “First Version of System, Architectural and Subsystem

Specification”

[2] VDI 4451 Blatt 6: 2003 – 01 Compatibility of automated guided vehicle systems (AGVS), Sensor systems for Navigation and Control. Berlin, Beuth-Verlag

[3] CommRob Deliverable D4.2.1 “Report on Haptic Based Detection”

[4] [Regenstein et al. 2007] (K. Regenstein, T. Kerscher, C. Birkenhofer, T. Asfour, J.M. Zöllner, R. Dillmann: Universal Controller Module (UCoM) - component of a modular concept in robotic systems. IEEE International Symposium on Industrial Electronics, 2007

[5] CommRob Deliverable D4.1. “Report on Human-robot interaction specification”