Neural networks-based terrain acquisition of unmarked area for robot mowers

Robot mowers are required to cover every part of the arbitrarily shaped lawn. In this paper the operational area populated with some obstacles without any manmade marks is managed differing from areas always well-marked. An effective terrain acquisition approach based on neural network is proposed for the robot mower. The robot circumnavigates the lawn and all obstacles to complete the terrain acquisition task under the proposed subsection acquisition strategy. The localization system with combined sensors serves the robot reliably to gain the digital boundary map of the unmarked operational area. According to the specialities of outdoor environment and the practical mowing requirements, the designed RBF network can make the accuracy of the terrain acquisition well improved and it is validated by experiment results. Finally the performance of the terrain acquisition is analyzed. The technical presentations in this paper can facilitate the development of the environmental robotics.