Real-time motion planning and execution of a 6DoF manipulator

In recent years, the desire to use manipulators in close collaboration with human workers has been a driving force behind a lot of robotics research. As part of a research project, Technolution BV has built a robotic pick and place system using the Universal Robots UR5, a 6 DoF manipulator. It was recognized that to realize the full potential of a robotic manipulator, it would have to be able to respond to unforeseen changes in the environment like dynamic obstacles or a moving target. Following these observations, the project was continued and aimed at researching methods that can be used to formulate a controller that is capable of adapting to a changing environment. A thorough analysis of the technical issues and theoretical background related to dynamic motion execution is presented in Part 1. Based on an examination of state of the art pick and place methods, a pick (or a place) motion is chosen to be dissected in two distinct parts: a global motion, which is used to move the end-effector to a pose close to the target object, and a Cartesian approach, to move towards the target object in a straight line and perform the actual grasp. For the global motion, a fast geometric path planner is a vital component. Several state of the art path planning algorithms are discussed and benchmarked with the specified scenario in mind. In order to ensure that a well performing configuration of each path planner is used, a novel tuning method based on automatic algorithm configuration tools was developed. An analysis of the performance of optimal planners was found to offer too few benefits for use in a system that requires fast reactions. To allow the Cartesian approach to be used when the target object is in motion, two different trajectory execution methods are tested. The first attempts to conform to existing infrastructure and uses cubic interpolation for determining a setpoint. This was found to offer accurate path execution and decent tracking performance. The 2nd method uses an on-line trajectory generation algorithm and was found to improve the tracking performance but did so using jerkier motions. For collision avoidance during the global motions, a time-scaling method has been combined with a replanning approach. Combining the global execution method with pose tracking, a system is proposed that is capable of responding to any changes in the environment