Planning of loco-manipulation tasks for a multi-arm space robot

Abstract

Significant developments of the space sector in recent years are now leading space technology on the brink of a new generation of large space structures. Those structures must be launched in pieces and deployed in situ. A key technology to allow large-scale costeffective on-orbit manufacturing and assembly is autonomous robotics. These concepts may bring relevant advantages for space missions in the near future, but cutting-edge motion planning tools are necessary to effectively operate robotic manipulators. The aim of this thesis is therefore to propose an approach for the motion planning of a robotic self-relocatable multi-arm system. The system is made of two arms mounted on a torso, summing up to a total of 15 degrees of freedom. It features three end effectors. Each end effector is equipped with a standard latching interface, that allows interaction with the world, both for locomotion and manipulation. The proposed approach consists of two layers of planning, a high-level planning of the contact sequence, and a low-level planning of the trajectories that perform the contact sequence. A final additional layer consists of the validation of the output. The motion planner provides plans for single locomotion and manipulation tasks, as well as combined loco-manipulation tasks. The contact planning is performed with numerical optimization and graph search methods. The path planning relies on kinematic tools and rapidly growing random trees. Eventually, trajectories are constructed with the lines and parabolas technique. The algorithms are tested in simulation. It is found that the developed motion planner works as desired in a great variety of cases. It handles different robots, environments and a large set of input tasks. The computing time is suitable for offline use. The outcome of the planning is a fully characterized motion plan, comprising instructions for all actuating devices of the system. It is locally optimal, simulation-validated and compliant with all the considered constraints of the problem.