Control Algorithms for a Shape-shifting Tracked Robotic Vehicle Climbing Obstacles

摘要

Research in mobile robot navigation has demonstrated some success in navigating flat world while avoiding obstacles. However, algorithms which analyze complex environments in order to climb three-dimensional obstacles have had very little success due to the complexity of the task. Unmanned ground vehicles currently exhibit simple autonomous behaviours compared to the human ability to move in the world. This research work aims to design controllers for a shape-shifting tracked robotic vehicle, thus enabling it to autonomously climb obstacles by adapting its geometric configuration. Three control algorithms are proposed to solve the autonomous locomotion problem for climbing obstacles. A reactive controller evaluates the appropriate geometric configuration based on terrain and vehicle geometric considerations. As a scripted controller is difficult to design for every possible circumstance, learning algorithms are a plausible alternative. A neural network based controller works if a task resembles a learned case. However, it lacks adaptability. Learning in real-time by reinforcement and progress estimation facilitates robot control and navigation. This report presents the reinforcement learning algorithm developed to find alternative solutions when the reactive controller gets stuck while climbing an obstacle. The controllers are validated and compared with simulations.