 |
International Society of Science and Applied Technologies |
| Intelligent Self-Healing Collaboration of UAV Swarms Based on Hawkes Processes and Deep Reinforcement Learning | ||||
| Author |
Zhongping Li
|
|||
| Co-Author(s) |
Qingan Qiu
|
|||
| Abstract | In complex mission environments, unmanned aerial vehicle (UAV) swarms are subject to random node failures and communication interruptions, rendering self-healing capability critical for the mission completion rate. This paper proposes an intelligent self-healing collaboration framework that integrates Hawkes processes with deep reinforcement learning. A mixed exponential kernel conditional intensity function with detection delay and fast–slow dual-mode decay is constructed, where an indicator function is employed to strictly guarantee zero excitation during the delay period. The self-healing collaboration problem is formulated as a Markov decision process, where the Hawkes conditional intensity serves as the core state and a deep Q-network (DQN) is utilized to learn the optimal self-healing timing and resource allocation. Simulations are conducted under various initial damage conditions, and the proposed strategy is compared with a threshold-based strategy and a no-action baseline. The results demonstrate that, compared with the threshold strategy, the proposed strategy improves the survival rate by approximately 39.7% on average and reduces the cumulative failure time by approximately 70.7%.
|
|||
| Keywords | UAV Swarm, Self-Healing Coordination, Hawkes Process, Deep Reinforcement Learning, Deep Q-Network | |||
| Article #: RQD2026-90 | ||||
Proceedings of 31st ISSAT International Conference on Reliability & Quality in Design |