doc. Kristian Hengster-Movric, Ph.D.

Autonomous and semi-autonomous vehicle formations are increasingly becoming a reality. Safe and fast coordination of these multi-agent systems thus presents a pressing issue, especially if operating in challenging environments. Distributed control, [1], and safety-critical control, [6], together offer means to address these problems. However, most instances in the literature for those purposes simplify the vehicle models to omnidirectional single or double-integrator dynamics, [7]. More accurate vehicle models are given by non-holonomically constrained mechanical systems, [4]. These are nonlinear systems, the coordination of which requires the use of the Internal Model Principle, (IMP), [2][3][5]. Virtually all existing results on the IMP-based cooperative control of heterogeneous nonlinear systems, however, assume oscillator-like systems with bounded states, e.g. [3],[5]. Vehicle models are significantly different in that their (position) states can theoretically attain unbounded values. Such cases require partial stability results, [8]. This research topic aims to address multi-agent safety-critical distributed control for realistic autonomous vehicle model formation keeping and fast reconfiguration in an unbounded environment, while ensuring collision and obstacle avoidance. Collision and obstacle avoidance can be achieved using potential field controls, [7]. It is a further challenge however to effectively integrate these potential field controls with the non-holonomic vehicle models under cooperative formation keeping protocols. Moreover, with the safe formation control of the closed-loop multi-agent system thus guaranteed, higher-level mission planning can be tackled by novel AI learning methods or MPC optimization, based on specific features of the considered theater of operations. The end-goal is to provide a realistic vehicle formation with a degree of autonomous agency while ensuring safety and agility in a variety of operational environments.