| Summary: | Human environments and tools are commonly designed to be used by two-handed agents. In order for a robot to make use of human tools or to navigate in a human environment it must be able to use two arms. Planning motion for two arms is a difficult task as it requires taking into account a large number of joints and links and involves both temporal and spatial coordination. The work in this thesis addresses these problems by providing a framework to combine two single-arm trajectories to perform a two-armed task. Inspired by results indicating that humans perform better on motor tasks when focusing on the outcome of their movements rather than their joint motions, I propose a solution that considers each trajectory's effect on the taskspace. I develop a novel framework for modifying and combining one-armed trajectories to complete two-armed tasks. The framework is designed to be as general as possible and is agnostic to how the one-armed trajectories were generated and the robot(s) being used. Physical roll-outs of the individual arm trajectories are used to create probabilistic models of their performance in taskspace using Gaussian Mixture Models. This approach allows for error compensation. Trajectories are combined in taskspace in order to achieve the highest probability of success and task performance quality. The framework was tested using two Barrett WAM robots performing the difficult, two-armed task of serving a ping-pong ball. For this demonstration, the trajectories were created using quintic interpolations of joint coordinates. The trajectory combinations are tested for collisions in the robot simulation tool, Gazebo. I demonstrated that the system can successfully choose and execute the highest-probability trajectory combination that is collision-free to achieve a given taskspace goal. The framework achieved timing of the two single-arm trajectories optimal to within 0.0389 seconds -- approximately equal to the time between frames of the 30 Hz camera. The implemented algorithm successfully ranked the likelihood of success for four out of five serving motions. Finally, the framework's ability to perform a higher-level tasks was demonstrated by performing a legal ping-pong serve. These results were achieved despite significant noise in the data. === Applied Science, Faculty of === Graduate
|
|---|
