Neural network control of mobile robot formations using RISE feedback.

Area of Science:

• Robotics
• Control Systems Engineering
• Artificial Intelligence

Background:

• Leader-follower robot formations require robust control strategies to maintain stability and avoid collisions.
• Approximating complex robot dynamics and ensuring asymptotic stability (AS) are key challenges in formation control.

Purpose of the Study:

• To develop an AS combined kinematic/torque control law for leader-follower robot formations.
• To utilize neural networks (NNs) with robust integral of the sign of the error feedback for approximating robot dynamics.
• To ensure formation stability and collision avoidance, even in the presence of obstacles.

Main Methods:

• Developed an asymptotically stable (AS) combined kinematic/torque control law using backstepping.
• Integrated a neural network (NN) with robust integral of the sign of the error feedback for dynamics approximation.
• Employed Lyapunov theory to prove AS of the formation and boundedness of NN weights.
• Analyzed formation stability during obstacle avoidance maneuvers.

Main Results:

• The proposed control law guarantees AS for the entire robot formation.
• NN weights are proven to be bounded, unlike typical uniformly ultimately bounded stability.
• Collision-free formation control is achieved, even when treating other robots as obstacles.
• Numerical simulations verified the theoretical stability conjectures during obstacle avoidance.

Conclusions:

• The developed NN-based control law ensures robust and AS leader-follower robot formation control.
• The approach effectively handles complex dynamics and guarantees collision avoidance during maneuvers.
• This method offers a significant advancement in multi-robot system stability and safety.