Navigation and control of an autonomous horticultural robot

Abstract Possible navigation schemes for an outdoor horticultural robot vehicle are considered. Problems encountered in applying techniques used successfully for industrial Automated Guided Vehicles are highlighted, with particular reference to the problem of finding vehicle position in an absolute Cartesian coordinate frame. An alternative navigation scheme is proposed, which relies only on local sensing capabilities. Demanded paths are specified without reference to absolute positions; instead the rows of crop themselves are employed as a navigational aid. Vehicle position is measured relative to the demanded path. A Kalman filter based algorithm is described which maintains an estimate of vehicle position in this path-relative coordinate frame; input is taken from a machine vision system which is capable of locating crop rows within a video image. This system permits guidance of an experimental vehicle along artificial crop rows at speeds of up to 1.5 ms−1 with an acuracy of ±20 mm.