Analysis, design and implementation of an optimal planner for redundant robotic applications

Modern robotic platforms for aerospace applications are complex systems characterized by many degrees of freedom. They are assigned with relatively simple tasks, defined at a high-level of abstraction. On the other hand, it is necessary that such tasks are optimized to save resources or increase the utilization of the expensive robotic assets. For such systems, globally-optimal redundancy resolution is a challenge, especially for those with more than one redundancy parameter. Among the leading techniques that are employed to solve the redundant inverse kinematics problem, approaches based on dynamic programming execute more efficiently by breaking the complex problem into various simpler subproblems and demonstrate superior flexibility in the management of constraints and objective functions. The present dissertation analyzes the globally-optimal kinematics inversion of mobile manipulators, i.e., systems composed of a manipulator mounted on a moving platform, and provides a new implementation for it. The newly developed algorithm is more flexible than previous technologies as it allows to manage multiple degrees of redundancy and is smoothly integrated with other tools required for space applications. The developments are based on ROS, the Robot Operating System, and its flagship planning library MoveIt!. The algorithm has been unit-tested and validated in simulation by using representative robotic platforms.