sensor placement by robot teams qiao li 7353963. outline concepts algorithms simulation analysis ...
TRANSCRIPT
Sensor Placementby Robot Teams
Qiao Li
7353963
Outline
ConceptsAlgorithmsSimulation AnalysisConclusionFuture WorkQuestions
Concepts
Sensor PlacementTypical Deterministic Deployments Pattern
s
Sensor Placement Sometimes sensors
are deployed in an unknown and/or hazardous environment, where physical dynamics
and spatio-temporal irregularities prevail.
There are two important
sensor placement methods, which do not need human
to place sensors in person: Carried-Based Deployment
and Sensor Self-Deployment.
R. Falcon, X. Li and A. Nayak, "Carrier-based Coverage Augmentation in Wireless Sensor and Robot Networks," M.S. thesis,Dept. Elect. Eng.,Univ. of Ottawa, Ottawa, Canada, 2011.
Sensor Placement Carried-based deployment involves mobile robots carrying
and dropping (static) sensors for optimal coverage formation. Sensor self-deployment deals with autonomous coverage formatio
n in WSN.
R. Falcon, X. Li and A. Nayak, "Carrier-based Coverage Augmentation in Wireless Sensor and Robot Networks," M.S. thesis,Dept. Elect. Eng.,Univ. of Ottawa, Ottawa, Canada, 2011.X. Li, H. Frey, N. Santoro and I. Stojmenovic, "Focused-Coverage by Mobile Sensor Networks" Univ. of Ottawa, Univ. of Padervorn, Univ. of Carleton, Ottawa and Paderborn, Canada and Germany, 2009.
Sensor Placement
Two requirements should be met in sensor deployment design for an area:
• Full Coverage: each point should be covered by at least one sensor.
• Connectivity: each sensor can be connected
to other sensors directly or through multiple hops
so that the data collected by individual node
can be relayed back to data sinks or controllers.
Z. Liao, J. Wang, S. Zhang and X. Zhang, "A Deterministic Sensor Placement Scheme for Full Coverage and Connectivity without Boundary Effect in Wireless Sensor Networks," Ad Hoc & WSN., vol. 19, pp. 327-351, May, 2012.
*Typical Deterministic Deployments Patterns
Hexagon ( )Square ( )Triangle ( )Srd 3
Srd
Srd 2
Algorithms
Snake-Like Deployment (SLD)Least Recently Visited algorithm (LRV)Back-Tracking Deployment (BTD)
Algorithms——SLD
Principle:• The robot moves step by step along a pre-computed geom
etric graph leading to optimal coverage,
each step to an adjacent empty vertex in the graph according to predefined rules for selecting moving direction, and drops a sensor after each step.
• SLD generates a snake-like S-shape robot trajectory.
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
Algorithms——SLD
Drawbacks:• SLD does not support multiple robots
or tolerate sensor failures.• SLD is very likely to be stuck at dead end
and leave uncovered areas in the environment,
thus it does not provide any guarantees
on the full area coverage even in the ideal failure-free environment.
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
Algorithms——LRV
Principle:• Sensors store weights for each direction that a robot
can travel in. The weights represent the number of times that a robot has visited each direction for a given sensor.
• The sensors then recommend the direction
with the lowest weight for the robot to travel
or the direction that is the least recently visited.
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
Algorithms——LRV
Drawbacks:• LRV requires many unnecessary movements
to fully explore a given Region of Interest (ROI).
These extra movements lead to an extremely large number of messages sent from the robot.
• It is unclear under what conditions LRV terminates.• LRV is used in the case of a single robot.
M. Batalin, G. Sukhatme, "Coverage, Exploration and Deployment by a Mobile Robot and Communication Network," Proceeding of the International Workshop on Information Processing in Sensor Networks., Palo Alto., USA, 2003, pp. 376-391.X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
Algorithms——BTD
Failure-Free Environment• Single-Robot Scenario• Multi-Robot Scenario
Failure-Prone Environment• Robot Failures• Sensor failures
BTD——Failure-Free Environment
Definition:• BTD was designed by equipping SLD
with an important back-tracking technique
and extent it to support multiple robots.
It was presented over a square grid graph.
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
BTD——Failure-Free Environment
Single-Robot Scenario:• Principle: The four geographic directions are pre-ordered
as West, East, North, South. This order defines preference when a robot selects its movement direction.
• Example:
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
BTD——Failure-Free Environment
Back-tracking:• A sensor colors itself white if it is adjacent to
an empty point and black otherwise.
It updates its own color dynamically.• The back pointer points to the location
of the first white sensor along the robot's backward path.• The robot moves to the next adjacent sensor
with the lowest sequence number
whose back pointer location is the same as
the robot's destination.
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
*BTD——Failure-Free Environment
Example:
BTD——Failure-Free Environment
Benefit:• The shortcut method achieves two goals:
a. It allows the robot to reach its destination
regardless of obstacles, using described retracing method.
b. It is also a more efficient method,
eliminating wasted movement by the robot.• For these reasons the shortcut mehod is employed
by BTD for robot back-tracking.
BTD——Failure-Free Environment
Multi-Robot Scenario:• Each robot follows the BTD algorithm as if it was the only robot
in the ROI. In a dead-end situation, if a robot can not find a back pointer on its current sensor, it will select a back pointer (if any exists)
stored in the neighborhood.• When a robot is back tracking to a white sensor, other robots
should not follow the same back-track path for that white sensor. Therefore during back tracking, the robot informs its current sensor
to erase back pointers along the forward path to the first encountered white sensor or the first sensor that stores no back pointer.
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
BTD——Failure-Free Environment
Example:
BTD——Failure-Prone Environment
Robot Failures (Easily Handled):• Before every movement step, a robot transmits
a ‘beacon’ message carrying its ID. If later
within a time window, its one-hop neighbors do not receive from it, these sensor nodes will consider that
the robot has failed and run a service discovery algorithm to find a nearby functioning robot to take over
the failed robot’s task.
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
BTD——Failure-Prone Environment
Sensor Failures (Difficultly Handled):• There are two steps in the fault-tolerance technique
to deal with their resultant sensing holes:
a. Find a search agent which is used to search
for the back pointer in a more efficient way.
b. Search for the back pointer and then move
to that back pointer to resume deployment.
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
Evaluation
Performance Metrics:• Coverage Ratio (CR): The average ratio of the number
occupied grid points to the total number of grid points.• Robot Moves (RV): The average number of movements
made by each robot during simulation.• Robot Messages (RM): The average number
of messages generated by each robot during simulation.• Sensor Messages (SM): The total number of messages
transmitted by sensors during simulation.
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
Evaluation
Modification on LRV:• For a fair comparison, LRV was slightly modified as follows:• Each sensor periodically transmits 'hello' message
as it does in BTD.• LRV was also extended to multi-robot scenarios.• As LRV does not terminate itself, it was also modified
to terminate as soon as each grid point has been visited
by a robot.
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
Evaluation
Simulation Setup:• A sensing hole is generated as follows:
the location of the first sensor to fail is selected randomly, then chose a random adjacent sensor to fail
until h sensors have failed.• Simulation environment:
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
Simulation Analysis
Coverage Ratio (CR):
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
Simulation Analysis
Impact of m on Movement and
Message Cost (RV/ RM/ SM): m refers to the number of robots
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
Simulation Analysis
Impact of h on Movement and
Message Cost (RV/ RM/ SM): h refers to the size of sensing
hole
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
Conclusion
BTD was demonstrated with both single-robot case
and multi-robot case in a failure-free environment
and a failure-prone environment. Simulation results indicate that BTD far outperforms
the only competing algorithms SLD and LRV
in various metrics.
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
Future Work
There are some parts in BTD algorithm can be improved in my future work:
• Sensing holes that do not affect robot back tracking
are not known by the robots and thus left untreated area in the failure prone situation.
• Cooperation between robots can be improved.
Questions
Question1: Please draw the square and triangle sensor deployment patterns. Hexagon is already given
as an example. (Tips: hexagon , square , triangle )
Srd Srd 2Srd 3
Questions Question2: Here is a brief description of the algorithm:
• The preference when a robot selects its movement direction is West>East>North>South.
• A sensor colors itself white if it is adjacent to an empty point.
• When the robot reaches a dead end,
it will back track to the nearest back pointer
(white sensor) along the backward path.
• The robot moves to the next adjacent sensor
with the lowest sequence number
whose back pointer location is the same as
the robot's destination. So could you please draw the possible path
for a single robot to deploy sensors
in the given area? (Start from A
and the black rectangle is an obstacle)
Questions
Answer:
Questions
Question3: Please try to proof the correctness of the following viewpoint:
• BTD terminates within finite time. Answer:• BTD terminates once all robots stop moving permanently.
A robot can visit a sensor no more than 4 times.
The number of sensor n is bounded, equal to the number of grid points contained in the ROI. Hence,
the maximum number of movements that a robot
can perform is 4n, implying each robot will make
finite number of moves between grid points.
This completes the proof.
Thank you!
BTD——Failure-Prone Environment
Find a Search Agent:• To handle the sensing hole problem, we introduce
a new color, gray. Sensors adjacent to a sensing hole color themselves gray.
• A robot identifies a sensing hole locally, as soon as
it finds that its back-tracking shortcut is broken due to loss of track of the back pointer and its current sensor
is a gray sensor.
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
BTD——Failure-Prone Environment
Example:
BTD——Failure-Prone Environment
Find a Search Agent:• The replacement sensors on the outer boundary of the pa
tch network learn back pointer information from adjacent gray sensors by listening to their 'hello' message.
• In order to resume back-tracking, the robot then sends
a search message in an arbitrary direction.• It finally stops, after a traversal of the outer boundary
of the patch network, at a directional extreme node.
This node is called search agent.
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
BTD——Failure-Prone Environment
Example:
BTD——Failure-Prone Environment
Search a Back Pointer:• The search agent sends a search message carrying
the back pointer along the border of the patch network.
This search message will erase the same back pointer
with a larger sequence number in the adjacent gray sensors; meanwhile, it will pick the location of the gray sensor
with smallest sequence number.
X. Li, G. Fletcher, A. Nayak and I. Stojmenovic, "Placing Sensors for Area Coverage in a Complex Environment by a Team of Robots," ACM Transactions on Sensor Networks., to appear.
BTD——Failure-Prone Environment
Example: