DESIGNING A MULTI-AGENT FRAMEWORK FOR UNMANNED AERIAL/GROUND VEHICLES

Abstract

This thesis focuses on developing a multi-agent framework that allows seamless usage of unmanned aerial or ground vehicle for the same mission. Multi-agent architectures for autonomous mobile robots are generally mission and platform oriented. Irrespective of the differences in aerial and ground robots, the algorithms for collision avoidance, path planning and path-tracking can be generalized. SOIFRA, developed in this thesis, is an interoperable multi-agent framework focusing on generalizing platform-independent algorithms for unmanned aerial and ground vehicles. Algorithms for collision avoidance, path-generation and path-following and human collision avoidance are incorporated into SOIFRA. Obstacle detection and avoidance are performed utilizing computer vision-based algorithms. VDPGT, developed in this thesis, is a platform independent real-time path-generation and path-following algorithm. A human visual system-based human detection and collision avoidance system is developed to add human collision avoidance behaviour into SOIFRA. Experiments performed utilizing multiple aerial and ground robots establish SOIFRA’s interoperability across several robotic platforms.