K-Means-Based Mobile Adaptive Networks for Swarm Robotic Systems with Non-Cooperative Behaviors

Mobile adaptive networks enable distributed adaptation and learning over networks while maintaining the cooperative motion, and play an important role in multi-robot control. Non-cooperative behaviors are common in swarm robotic systems and can greatly affect the system performance through information transmission and fusion in mobile adaptive networks. Some adaptations of different aggregation rules have been emerged to improve fault tolerance in mobile adaptive networks. This paper proposes K-means-based mobile adaptive networks which use K-means clustering algorithm to detect the faulty neighbors with non-cooperative behaviors and carry out mobile adaptive networks depending on information of the normal neighbors. Simulations are carried out in four existed mobile adaptive networks with different aggregation rules and our K-means-based mobile adaptive networks with these four aggregation rules. Simulation results prove that our proposed algorithms have better fault tolerance than the original algorithms.