Multi-objective path planner for an agricultural mobile robot in a virtual greenhouse environment

Robotics in agriculture has experienced enormous development in the past decade. In order to accomplish numerous agricultural tasks, the order in which the crop is covered is particularly important in order to minimise travelling cost and to preserve soil conditions. However, it is difficult to determine an optimised sequential path for the robot which minimizes navigation costs, while ensuring a fully completed agricultural operation. This paper presents the implementation of a multi-objective algorithm to solve the path planning problem for pesticide spraying operation inside a greenhouse. Pesticide spraying problems found in greenhouses have been adapted from the vehicle routing problem found in operations research, where the infected plants represent the customers and the mobile robot represents the vehicles. Routes between the plants are generated using a Probabilistic Roadmap path planner based on the designed virtual environment. The virtual environment is designed based on the real greenhouse environment to visualise the agricultural operation inside the greenhouse. To determine the best routes for the mobile robot, the Non-dominated Sorting Genetic Algorithm using Reference Point Based (NSGA-III) is tested and applied to the system. The solution quality has been compared with the Non-dominated Sorting Genetic Algorithm (NSGA-II) using the C-metric indicator. Comparisons with NSGA-II using the C-metric indicator verify that NSGA-III offers a superior performance with a good quality result.