Abstract
In the robot swarm, each robot can freely form swarm with others to share information. Although particle swarm optimization (PSO) has been demonstrated to outperform Q-learning and evolutionary algorithms, less study is conducted to characterize various swarm intelligence (SI) algorithms and evaluate their performances for robot swarm learning. In this research, we select three representative SI algorithms include bat algorithm (BA), PSO, grey wolf optimizer (GWO) according to their learning strategies. These three algorithms are implemented in a distributed manner and compared under various number of robots (NR) and communication ranges (CR). The simulation results demonstrate that: 1) PSO outperforms BA and BA outperforms GWO in general, 2) GWO performs better than PSO and BA under large NR and long CR, 3) increasing NR and CR can significantly improve the performance of GWO. These results can shed lights on selection of optimal SI algorithms for robot swarm under various scenarios.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have