Orchard traveling UGV using particle filter based localization and inverse optimal control

The authors previously proposed an Unmanned Ground Vehicle (UGV) in an orchard as a base platform for autonomous robot systems for performing tasks such as monitoring, pesticide spraying, and harvesting. To control a UGV in a semi-natural environment, accurate self-localization and a control law that is robust under large disturbances from rough terrain are the first priorities. In this paper, a self-localization algorithm consisting of a 2D laser range finder and the particle filter is proposed. A robust nonlinear control law and a path regeneration algorithm that the authors proposed for underactuated mobile robots are combined with the localization method and applied to a drive-by-wire experimental vehicle. Excellent experimental results were obtained for traveling through a real orchard. The standard deviation of the control error in the lateral direction was less than 15cm.