Distributed control for flock navigation using nonlinear model predictive control

Abstract

We present a novel approach that employs nonlinear model predictive control (NMPC) to address flock control while incorporating obstacle avoidance within a leader-follower framework. This strategy utilises each agent's predictive capabilities to navigate effectively. Building upon Reynolds' foundational flocking principles of cohesion, separation, and alignment, we adapt them to serve a navigation objective rather than focusing solely on formation tasks. In doing so, we introduce novel concepts that enhance our approach's robustness and flexibility. These concepts include the assessment of information credibility and significance received from neighbouring agents, along with the dynamic adjustment of trade-offs between reference values. We incorporate these features into our NMPC formulation and illustrate their advantages through numerical simulations and laboratory experiments.