Robust Time-Varying Formation Control of Heterogeneous Multi-Agent Systems Using Novel Potential Field Avoidance

Abstract

This thesis investigates time-varying formation algorithms for teams of heterogeneous multi-agent systems consisting of ground based two-wheel mobile robots and aerial quadrotors while developing novel obstacle avoidance algorithms to improve the navigation capabilities of the agents. Sliding mode controllers are developed based on designs from literature, with some refinements for their specific applications. These are tested for a variety of cases in order to ensure they will be sufficient for the control of agents in formation. The obstacle avoidance algorithm developed is an improvement on the artificial potential field method. By using additional information about the position of the agent relative to an obstacle, a more informed obstacle avoidance protocol is developed, which reduces the impact of the repulsive fields of the obstacles. This is done primarily based on factors such as the heading of the agent relative to the obstacle. Two methods for time-varying formation are then developed which use the previously developed controller for navigation, and the obstacle avoidance to prevent collisions. With these algorithms, agents navigate through an environment by making collective changes in their formation based on the information about the environment gathered by leader agents. Simulated and experimental testing validates the effectiveness of these methods.