On June 4th, the Hujiang Economics and Management Forum Series (112nd Session) was successfully held in the second conference hall on the fourth floor of Building A of the Economics and Management Building. Associate Professor Liu Lei delivered an academic lecture titled Research on Reinforcement Control of Micro Cluster Robots Inspired by Fish School Movement.
The lecture introduced a novel approach that combines the fish school attention mechanism with deep learning and reinforcement learning to establish an attention-based reinforcement collaborative control model. This model was then applied to the control of cluster robots. The proposed attention-based fish school collaborative control model and the method of transferring biological clustering models based on deep reinforcement learning were validated through simulation experiments. These experiments demonstrated the model's and method's effectiveness. The fish attention mechanism focuses on the critical information of at most two key neighbors, constructing an attention-based fish school collaborative interaction model. This model shows excellent decoupling capabilities and superior clustering performance.
To verify the simulation results, the Cuboids robot was designed and developed. This robot features modularity, enhancing its adaptability and flexibility, with more robust performance to ensure stable operation under different environmental conditions. Experimental results indicate that the attention-based fish school collaborative control model retains the robust flexibility of biological clusters while achieving task controllability for cluster robots, enabling them to complete predetermined tasks efficiently.
Translated by Liu Lei
Reviewed by Fang Zhiming
