Aiming at the low efficiency, high risk and high cost in manual ship descaling, the wallclimbing robot is developed. This paper adopts the distributed control scheme of host and slave machine to research the controller of the wallclimbing robot based on ARMLinux, including the sensor, the comparator, and the composition of the circuit. By using the reinforcement learning algorithm QLeaning algorithm, the reinforcement learning tracking of wallclimbing robot is realized. The weakness that traditional PID and other algorithms that can't optimize the action strategy for environment are improved and the wallclimbing robot in different environments tracking accuracy is raised. The experiment results show that the control system of the wallclimbing robot based on ARMLinux can perform better and meet the requirements of the controller.