Distributed Optimal Formation Tracking for Heterogeneous Open Swarm Systems With Dynamic Identities: Theory and Experiment

Le résumé fourni par la source

This article addresses discrete-time optimal formation tracking problems in heterogeneous open swarm systems, where members may join, leave or rejoin dynamically, and the total population is neither fixed nor known a priori. The system is vulnerable to time-varying cyber attacks that can dynamically alter agent identities between honest and adversarial states. Each honest agent seeks to track time-varying optimal trajectories by minimizing the sum of local objective functions over all honest agents during its active period, while maintaining predefined formation configurations. A distributed algorithm is proposed, enabling agents to estimate local subgradients from bandit feedback, and track optimal trajectories using neighboring estimates without identity knowledge. Convergence analysis demonstrates that each honest agent achieves sublinear growth of dynamic regret over time. Simulation and experimental results of a swarm system, involving five unmanned ground vehicles and two unmanned aerial vehicles, validate the theoretical results.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.