Neural Network-Cascade UDE Based Tracking Control of Leader-Follower Agents with Compound Uncertainties
Tong Li, Weihao Li, Zhiqiang Li, Mengji Shi, Boxian Lin, Zixuan Liu · 2021
This paper concerns with the multi-agent distributed leader-follower tracking control problems in the presence of compound uncertainties. These unknown time-varying parts not only lie in the both integral units of the second-order follower dynamics, but also show up in the leader’s kinematics. A cascade uncertainty and disturbance estimator (CUDE) based backstepping controller is designed for the followers, which is capable of handling the model uncertainties and input disturbances that respectively appearing in their first-and second-order integral units. However, UDE based methods require accurate modeling of the control system, while the leader is not controllable. Thus, for the leader, the uncertain parts are compensated by constructing a neural network (NN) estimation mechanism. Adaptive estimation parameters are designed according to Lyapunov stability theories to enhance the robustness of the system. The combination of NN and CUDE requires less computational resource than conventional NN-based methods while enabling a valuation of the uncertainty in the leader’s equation. Finally, a numerical simulation is conducted to verify the effectiveness of the control scheme.