Smart agriculture: Development of a skid-steer autonomous robot with advanced model predictive controllers

Abstract

The agricultural domain has been experiencing extensive automation interest over the past decade. The established process for measuring physiological and morphological traits (phenotypes) of crops is labour-intensive and error-prone. In this paper, a mobile robotic platform, namely The Autonomous Robot for Orchard Surveying (AROS), was developed to automate the process of collecting spatial and visual data autonomously. Furthermore, six different control frameworks are presented to evaluate the feasibility of using a kinematic model in agricultural environments. The kinematic model does not consider wheel slippage or any forces associated with dynamic motion. Thus, the following six controllers are evaluated: Proportional-Derivative (PD) controller, Sliding Mode Controller (SMC), Control-Lyapunov Function (CLF), Nonlinear Model Predictive Controller (NMPC), Tube-Based Nonlinear Model Predictive Controller (TBNMPC), and Model Predictive Sliding Mode Control (MPSMC). This paper provides insight into the degree of disturbance rejection that the mentioned control architectures can achieve in outdoor environments. Experimental results validate that all control architectures are capable of rejecting the present disturbances associated with unmodelled dynamics and wheel slip on soft ground conditions. Additionally, the optimal-based controllers managed to perform better than the non-optimal controllers. Performance improvements of the TBNMPC of up to 209.72% are realized when compared to non-optimal methods. Results also show that the non-optimal controllers had low performance due to the underactuated constraint present in the kinematic model.