Safe, Resilient and Efficient Operation of Autonomous Aerial and Ground Vehicles

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

Autonomous vehicles working alone or in coordination with other autonomous vehicles have become indispensable in recent years for many military and law enforcement missions (e.g., in surveillance, long-range communication, target acquisition and tracking, and even weapon delivery). Most recently, the aerial versions of these autonomous vehicles ('flying robots' or 'drones') have entered the civilian sector and have been used successfully in many applications such as fire detection, crop dusting, aerial surveying, entertainment industry, traffic monitoring, infrastructure inspection, weather/hurricane monitoring, and commercial product delivery, to name a few. The unmanned air vehicle (UAV) market is exploding and most often than not these autonomous vehicles will have to operate in an environment that is highly uncertain, and even adversarial. For instance, for the case of unmanned aerial or marine/underwater vehicles winds or sea currents have a great impact on system performance. Similarly, self-driving vehicles must operate and interact with the surrounding traffic flow. The theory and methodologies developed in this research will make it possible to enable better coordination of such autonomous vehicles and systems, thus increasing their efficiency, reliability, and overall performance. The results of this research will help in extending the usability and endurance of aerial and marine autonomous vehicles and contribute to the safe operation of self-driving vehicles in traffic. The educational aspects of the project also includes efforts for involving minority and other under-represented students in the research. The research tackles a fundamental problem in the area of differential games and optimal trajectory generation for autonomous vehicles operating in the presence of exogenous or endogenous disturbances. Recent advances in the analysis of coordinated control of multi-agent systems in the presence of external flow fields using Voronoi-like decompositions, along with numerical techniques based on level sets will be utilized to solve multi-agent pursuit-evasion and target assignment problems in a numerically efficient manner. A novel reachability set inclusion property allows for the solution of a large class of such pursuit-evasion problems with multiple agents, even under the influence of external disturbances such as the drift field. The results of this research will allow coordination strategies and distributed pursuit-evasion protocols that go beyond the usual Euclidean metrics to establish proximity relationships between the agents, including time, energy and fuel. Extensions to stochastic environments will also be addressed, by leveraging new decomposition methods to solve stochastic partial differential equations arising from a stochastic formulation of level set propagation.