Multiobjective path optimization for autonomous land levelling operations based on an improved MOEA/D-ACO

Abstract

To improve the efficiency of the land levelling operation of a tractor-scraper system controlled by an autonomous guidance system which depends on the complete earthwork transportation and minimum navigation cost, it is crucial to optimize the travelling paths. This article presents an innovative method using an improved multiobjective algorithm to solve the path optimization problem for autonomous land levelling operations. First, an environment model was built based on collected field data, and a model decomposition method was proposed to decompose the field terrain model and realize complete land levelling. Then, a three-objective problem model (consisting of the travelling distance, steering angle and earthwork) was developed, and the Pareto set of the travel sequence was obtained using a Multiobjective Evolutionary Algorithm Using Decomposition and Further Mutation Ant Colony (MOEA/D-FMACO). Finally, the path optimization of five field cases based on three algorithms, the MOEA/D-FMACO, MOEA/D-ACO and NSGA-III, was simulated in MATLAB. The simulation results show that the MOEA/D-FMACO has better solution performance quality compared with the other two algorithms and shortens the execution time. In field experiments, a GNSS-based land leveller controlled by automatic steering and levelling systems tracked the planned paths based on the MOEA/D-FMACO and a previous method for a field in two periods. The field results verified that the land levelling effect based on the optimization paths of the MOEA/D-FMACO had a significant improvement over the previous method. The proposed method decreased the maximum elevation difference (ΔZ) by 10.5 cm and the flatness of the field (Sd) by 4.3 cm and increased the elevation distribution around the target elevation (η) by 11.15% compared to the previous method.