ROS-based Control Of Reconfigurable Modular Robots

Re-configurable modular robots(RMRs)support flexible changes of kinematic configurations according to the task at hand.As the mechatronic systems are constantly improved,challenges of software development for RMRs remain,which include: 1)Flexibility: The software system should be able to deal with all possible kinematic configurations instead of only a few specific ones;2)Ease-to-use: This is a precondition for applying the modular robots for field tasks;3)System openness: which allows further development by system integrators,e.g.adding sensors;4)Standardization: It shortens the development cycle and eases the maintenance process.Facing these challenges,this thesis introduces a software framework based on ROS(Robot Operating System)for our re-configurable modular system.Meanwhile it tries to answer this question: how to utilize the ROS framework and its tools to develop for RMRs a software system which applies to all possible RMR kinematic configurations instead of some specific ones.With this in mind,the paper mainly discusses the following three aspects:1.Using URDF/Xacro format to describe all module types,the built modular robot can be expressed by specifying all constituent modules and the connection between different modules.Besides,other configuration files can also be generated automatically according to the kinematic structure.2.With each module attached with QR markers,modules that constitute a robot are identified using an external camera.The whole kinematic structure can then be identified using geometric constraints between connected modules.The robot model is generated automatically and visualized through RViz.3.CANopen communication between the ROS universe and the real robot are achieved by using an open-source package: ros_canopen.Then three basic robot control mode including joint-space control,end-effector Cartesian-space control and offline generated data replay are tested with a modular manipulator as an example.The generic kinematic solver TRACK-IK,which employs a numerical approach,is used so that these codes can be applied to other modular robots of different kinematic structures as well.The goal of this thesis is to develop a configuration-agnostic control framework using ROS mechanisms,while configuration-related information is automatically generated with template files.In doing so,the configuration variety is encapsulated in the system,which lowers the barrier for users to develop their own applications.