A Modular, Behaviour-Based Hierarchical Controller For Mobile Manipulators

Abstract

A mobile manipulator is a robotic system consisting of a robotic manipulator mounted onto a mobile base. This greatly extends the workspace of the robotic manipulator and allows it to perform more tasks. However, combining both systems increases the complexity of the control task as well as introducing additional controller tasks such as coordination of motion, where executing the task can involve using both the mobile base and manipulator, and cooperation of task, where many tasks can be executed at once. In this thesis a controller for a mobile manipulator is developed from smaller, simple controller blocks, allowing the controller to be flexible, easy to understand, and straightforward to implement using well-known embedded software implementation approaches. A behaviour-based approach was used to build the individual controllers, and a hierarchical structure was used to organise the individual controllers to provide cooperation between them and coordinated motion. The task assigned to the controller was to reach a series of waypoints in a large workspace, while satisfying performance metrics of manipulability and tip-over stability. The operation of the controller was tested in simulation using 100 randomly generated scenarios consisting of five randomly generated waypoints in each. Using default thresholds for manipulability and tip-over stability, the controller was successfully able to complete all scenarios. Further simulations were then performed testing the effects of varying the thresholds of the performance metrics to explore the tradeoffs involved in different parameter choices. The controller was successful in a majority of these scenarios, with only a few failing due to extreme threshold choices. The reasons for these failures, and the corresponding lessons for robot designers are discussed. Finally, to demonstrate the modularity of the controller, an obstacle avoidance controller was added and simulation results showed the controller was capable of avoiding obstacles while still performing the same tasks that were used in previous tests. Successful simulation results of the controller across a range of performance metrics shows that the combination of a behaviour based and hierarchical approach to mobile manipulator control is not only capable of producing a functional controller, but also one that is more modular and easier to understand than the monolithic controllers developed by other researchers.