The Application of GD-kWTA Network in Multi-Robot Competition

This paper applies the gradient-based differential kWTA (GD-kWTA) algorithm to multi-robot tracking tasks based on the principles of multi-robot competition and collaboration. At the same time, the kWTA network is modeled, and an activation function is added under noise-free conditions to improve the network's convergence speed. In the kWTA numerical experiment, special nonlinear activation functions are designed to activate the kWTA network to verify that the activation function has exponential convergence. Finally, under limited communication conditions, a consensus filter and the GD-kWTA network are combined as a robust control strategy to perform multi-robot target tracking tasks. At the same time, complex motion trajectories are set to further verify the effectiveness and reliability.