Data-Driven Active Disturbance Rejection Control of Plant-Protection Unmanned Ground Vehicle Prototype: A Fuzzy Indirect Iterative Learning Approach

Abstract

This letter proposes a fuzzy indirect iterative learning (FIIL) active disturbance rejection control (ADRC) scheme to address the impact of uncertain factors of plant-protection unmanned ground vehicle (UGV), in which ADRC is a data-driven model-free control algorithm that only relies on the input and output data of the system. Based on the established nonlinear time-varying dynamic model including dynamic load (medicine box), the FIIL technology is adopted to turn the bandwidth and control channel gain online, in which the fuzzy logic system is used to update the gain parameters of iterative learning in real time. Simulation and experiment show the FIIL-ADRC scheme has better control performance.With the advancement of agricultural intelligence, the artificial spraying method that threatens health is being replaced by the plant protection UGV. The speed and steering angle control of plant protection UGV is particularly important in the process of field operation. The accuracy of steering angle determines whether the UGV needs to be manipulated artificially at the boundary corner of the field [1]. In addition, with the spraying of the liquid, the total mass of the vehicle will be reduced, and the sloshing of the liquid in the medicine box, air resistance, nonlinear friction and the unmodeled part of the system will cause multi-source and unknown interference to UGV. Based on the conditions above, the model of the plant protection UGV is hard to establish. When the model cannot be precise enough, the control effect cannot be further improved, data-driven control schemes can solve this problem [2], [3]. However, the general data-driven control approaches are mostly based on error elimination control methods. When there are unknown external disturbances and internal uncertainties, they cannot predict disturbances well to provide control compensation, and the control effect cannot meet the ideal control requirements. Compared with the general datadriven control approaches, the advantage of ADRC is great when dealing with unknown disturbance, which is more suitable than other model-free data-driven control methods. Therefore, ADRC scheme, a model-free data-driven control method, is used in this letter as basis to control plant protection UGVs. The core idea of ADRC scheme is to treat uncertain factors as total disturbance, and estimate the total disturbance online by the extended state observer (ESO), and compensate the control input [4]. At present, the research on ADRC of UGVs has made some achievements [5].