Wheel slip modeling, estimation and control for AGVs in agricultural applications

Automated guided vehicles are an emerging technology that has been proven to increase prodictivity in warehouses and manifacturing, most often by automating transportation processes. Octinion, an R&D compa- ny specialized in mechatronic product development applied to biological material, exploits AGVs technology to automatize greenhouse mainten- ance. This thesis focuses on improving the performance of Xenion, an AGV developed by Octinion that suffers from wheel slip while cornering. Given the absence of a feedback controller, the AGV is controlled in an open loop fashion, which leads to a low accuracy of the vehicle’s trajectories, since in certain conditions the vehicle model does not repre- sent accurately enough the real vehicle dynamics. In order to accurately model wheel slip and improve the accuracy of the vehicle, a detailed dynamic vehicle model including all the non idealities of Octinion’s AGV has been developed. Combining the measures obtained from a camera-based motion capture system, with the AGV’s control inputs, an optimal search and a grid search method have been used to perform an offline parameter estimation for the newly developed vehicle model. This led to a decrease in the error of the predictions of the final position of the robot after cornering trajectories by over 80%. Finally, given the complexity of the developed vehicle model, an optimal motion plannig approach has been adopted to find an optimal control strategy for the vehicle.