Vehicle IntelligenceSWARMED VEHICLES THE FUTURE OF TRANSPORTATION

Narendar.R.CVIT University

Introduction The combination of increasing population, growing demand for mobility, and limited capital and real estate for new roads strains the current transportation infrastructure and constrains future options.

Driven by this situation, and spurred by the convergence of key technologies in multiple domains, the emergence of a “swarm” concept of operations for automotive transportation is possible, perhaps likely.

Flying SWARM of UAVs

Swarm Vehicles We define a swarm in this context as a collection of vehicles traveling at high speeds and in close proximity using passive and/or active cues to determine behaviour.

Potential technologies required for swarm transportation include peer-to-peer Networking enabled by Wi-Fi communication, precision GPS navigation, advanced sensors, and more agile drivetrains.

Swarm has many technologies embedded it to make the vehicle intelligent enough. Even various algorithms are being developed and each are proprietary with the vendor.

Optimization for Shared-Autonomy in Automotive Swarm Environment Sue Bai - Honda R&D Americas, Inc.

Aaron Sengstacken and Daniel DeLaurentis - Purdue University

Need For Swarm Intelligence The benefits obtained from a realized swarm-type operation in terms of reduced fuel consumption and pollutants as well as increased traffic throughput are expected to be significant.

They have found that fuel consumption and pollution levels were reduced during rapid acceleration transients by 28.5% and 1.5%-60.6%, respectively, in the presence of only 10% semi-automated vehicles.

It also improves the overall transportation economy and accident prevention.

Objective Effective techniques for managing the swarm are needed to ensure safe, efficient, and reliable interactions between vehicles. Our overarching research goal is to synthesize optimal control architectures for the development of “automotive intelligence” that enables effective management and safe participation in a swarm architecture under a wide variety of operational scenarios.

Sensing•Proximity•GPS•User Steer

Artificial Intelligence•Communication

•Swarm Intelligence

Dynamic Control•Model Driven Control

•Human Input•Cognitive Inference

Control LogicThe control logic for each vehicle containing the four agents employed an “outer loop - inner loop” control hierarchy.

1. The human reactive agent represents the control logic and input produced by a human driver. Human reactive agent has limited large-scale manoeuvrability and incorporates a time delay to represent the slow reaction times of human drivers.

2. The information agent represents a suite of environmental sensors that process and perceive potentially hazardous situations.

3. The machine reactive agent represents the vehicle autonomy, and therefore the relation between the machine and human reactive agents provided the means to examine different shared- autonomy approaches.

The Paradigm Shift The present day architecture for vehicle-to-vehicle interaction emphasizes the human driver, with minimal vehicle autonomy used for detection of problems and collision avoidance warnings.

Effective techniques for managing the swarm are needed to ensure safe, efficient, and reliable interactions between vehicles.

We avoid both extremes of vehicle control—fully autonomous and human driver-exclusive operations—seeking instead to augment the human driver’s intelligence through a shared autonomy approach.

Human ComponentMachine Component

SHARED AUTONOMY CONTROL The goal of a shared autonomy control approach is not to replace the human driver in ground transportation but instead to optimally augment the human’s capability and intelligence.

We seek to utilize the strengths of the human driver and supplement his/her weaknesses with levels of vehicle autonomy. The shared autonomy model is not itself intelligent; rather intelligence is a result of the cooperative behaviour of multiple vehicles following simple reactive rules.

Instinct Cognitive Thinking

Fast Computation

of Sensor Data and Decision

making

Shared Autonomy

Limitation Human Perception The distinguishing features of the human driver are strength in high level reasoning, with limited sensory perception and slow reaction times.

The information agent is equipped with a range of environmental sensors and augments human perception by alerting the driver of potential upcoming hazardous situations.

ArchitectureShared Autonomy

Human

Cognitive Agent Reactive Agent

Machine

Information Agent

Machine Reactive Agent

Machine Reactive Features Obstacle Avoidance Behaviour:

Obstacle avoidance behaviour prevents collisions with static obstacles as well as other vehicles. The obstacle avoidance behaviour was a multiple input multiple output fuzzy rule set to control both the steering angle and speed.

Path Tracking Behaviour:

The path tracking behaviour within the human cognitive agent was goal oriented. Desired paths were tracked using waypoint navigation . Path tracking is a multiple input multiple output behaviour to control both the steering angle and vehicle speed.

Vehicle swarm abstraction:

Interacting agents with their spheres of influence

Optimization Setup Optimization of the shared-autonomy swarm operation was performed using a genetic algorithm (GA).

GAs are useful for pinpointing preferred swarm architecture options as well as parameters within the fuzzy rule set .

Hence, they provide an appropriate framework for optimizing unknown parameters of the fuzzy rule base system and the architecture of cooperation among the vehicles participating in the swarm.

The Algorithm the basic four steps used in simple Genetic Algorithm to solve a problem are,1. The representation of the problem2. The fitness calculation3. Various variables and parameters involved in

controlling the algorithm4. The representation of result and the way of

terminating the algorithm

Swarm Based Consumer Vehicles

Research prototype of the Audi series car being tested fro swarm intelligence by the Volkswagen automotive innovation laboratory.

MicroMAX

The first commercial vehicle that is fully capable of a swarm based utility that can render a driverless approach towards any traffic entity.

Inclusive of all path tracing and obstacle avoidance technologies.

References1. Optimization for Shared-Autonomy in Automotive Swarm Environment Sue Bai - Honda R&D

Americas, Inc. Aaron Sengstacken and Daniel DeLaurentis - Purdue University :: SAE Journal ::;2. E. Bonabeau, M. Dorigo, and G. Theraulaz, Swarm Intelligence: From Natural to Artificial SystemsNew

York: Oxford Univ. Press, 1999. 3. A. Bose and P. Ioannou, “Analysis of traffic flow with mixed manual and semiautomated vehicles,” IEEE

Trans. Intell. Transp. Syst., vol. 4, pp 173-188, Dec. 2003. 4. C. Ye, N. Yung, and D. Wang, “A fuzzy controller with supervised learning assisted reinforcement

learning algorithm for obstacle avoidance,” IEEE Trans. Syst., Man, Cybern. B, vol. 33, pp. 17-27, Feb. 2003.

5. J. Wu, T. Liu, “Fuzzy control stabilization with applications to motorcycle control,”IEEE Trans. Syst., Man, Cybern. B, vol. 26, pp. 836-847, Dec. 1996.

6. J. Ackermann, P. Blue, T. Bünte, L. Güvenç, D. Kaesbauer, M. Kordt, M. Muhler, and D. Odenthal, Robust Control: The Parameter Space Approach. London, U.K.: Springer-Verlag, 2002.