SwRI has developed a flexible behavior architecture to enable unmanned vehicles (air or ground) to cooperatively patrol a defined perimeter with no central controller, and to effectively respond to detected anomalies based on the number and type of vehicles in the system. Agent-based modeling and simulation methods were used to develop and test these algorithms, including the message protocol that allows the shared integration and representation of anomaly data.
SwRI has developed a model to capture the behavior of individual unmanned ground vehicles (UGV) performing a collective mission of perimeter patrol. Each vehicle has individual capabilities related to sensor field of view (FOV) and path planning, as well as the ability to communicate with other vehicles in the system. When an anomaly is detected, which might be an object blocking the vehicle’s path or an unexpected moving object, each vehicle decides to respond or not based on several factors, including how many other vehicles are already responding. The vehicles can then request updates on the status of the anomaly to re-evaluate their individual responses. Using this strategy, the vehicles can effectively coordinate a response to one, or more, anomalies while maintaining an acceptable level of coverage over the entire perimeter. SwRI has begun deploying and testing these behaviors using the unmanned ground vehicles in our inventory, and is investigating the addition of unmanned aerial vehicles to the system.
Key Elements of SwRI’s Coopertive Perimeter Patrol Research:
- Flexible behavior architecture for multiple vehicles under decentralized control
- High-fidelity micro-simulation of unmanned ground vehicles performing a cooperative perimeter patrol mission
- Integration with commercially-available software tools
- Extensibility to aerial vehicle systems
SwRI has developed a high-fidelity micro-simulation of UGVs performing a perimeter patrol mission, which cooperatively respond to detected anomalies without a central controller. The simulated vehicles have perception and navigation capabilities that are representative of SwRI’s UGV fleet. The core behavior architecture for the cooperative response to anomalies enables the system of UGVs to adapt in near real-time to changing conditions, such as the detection of multiple anomalies. Individual vehicle behaviors also prevent the situation where too many vehicles have responded to an anomaly, leaving insufficient surveillance around the perimeter, by monitoring the collective response of the system, and either joining or leaving the response as appropriate.
- Avery, P.A. Learning and Adaptation in Cooperative Vehicle Systems. Presented at the 2013 Society of Automotive Engineers of Japan (JSAE) Annual Congress, Paper presented, Yokohama, Japan, May 2013.