Safety-Critical Multi-Agent Flocking via Motion-Aware Control Barrier Functions

This paper presents a safe and scalable control framework for heterogeneous multi-agent flocking using motion-aware control barrier functions. A nominal flocking controller is designed for cohesion and velocity alignment, while inter-agent separation and collision avoidance are enforced through pairwise safety constraints in local quadratic programs (QPs). Agent heterogeneity is explicitly considered, allowing for different physical radii and actuation limits. To reduce conservatism and computational burden, explicit reaction radii are derived for a distance-only safety activation policy, which characterise when pairwise safety constraints are automatically satisfied and can be removed from the QP. A motion-aware activation policy is then proposed that exploits both relative distance and approach velocity to activate safety constraints only when there is a potential collision risk. Simulations validate the effectiveness and scalability of the proposed framework, while real-world Crazyflie experiments demonstrate its practical feasibility.