Vision-Based Distributed Heterogeneous Swarm Navigation Method

The centralized structure commonly used in the current heterogeneous swarm navigation methods is unreliable, has poor scalability, and relies excessively on communication and external location information. A vision-based distributed heterogeneous swarm navigation method is proposed in this paper, a deep convolutional neural network is adopted to detect nearby agents, the accuracy and speed of detection is enhanced through data enhancement and model pruning. It is shown from simulation experiments that the heterogeneous swarm does not rely on communication and external location information, but only relies on vision, which can achieve safe navigation and formation maintenance.