Efficient motion planning for mobile manipulation in planetary exploration

Abstract

Space is the main frontier to further expand our human civilization and our knowledge of the universe. Exploring space is challenging and expensive since humans are not biologically prepared to survive in the harsh conditions of space. Robots are a suitable alternative though, since they can reach and explore space in a much cheaper way and without risking human lives. In the last decades the so-called planetary exploration vehicles, or rovers, have been able to gather very interesting information about the Moon and Mars. They do so through scientific instruments, which are generally held by a robotic arm that places them in scientifically interesting targets on the surface. Rovers find a main drawback in remote teleoperation from Earth, given the huge challenges of transmitting information in space (delays, limited communication windows). Therefore, the scientific return of planetary missions is greatly increased if the rover can perform the task autonomously, i.e. without the need for human intervention. Considering the mobility provided by the robot’s locomotion system and the manipulation capabilities of the robotic arm, planetary rovers can be viewed as mobile manipulators. Since a great amount of the scientific tasks that a rover performs include mobile manipulation movements, autonomously executing them would raise the scientific return of the planetary exploration mission. Though autonomous navigation on Mars is already quite advanced, demonstrated for instance with the Perseverance rover, autonomously performing mobile manipulation tasks is still troublesome. Such a challenge is mainly caused by the complexity of planning and controlling the movements of the mobile platform and the robotic arm together, which cannot be easily achieved in space due to the limited computational resources available out of the Earth. Hereby, this doctoral thesis presents the research conducted by the author to achieve autonomous mobile manipulation in planetary exploration vehicles. The goal is to plan and control the motion of the robot efficiently, i.e. with low computational cost, to perform the mobile manipulation scientific tasks. To do so, on-Earth technology has been analyzed, modified, and applied to the field, with four main contributions. First, the global path planner methodology to reach the area of scientific interest is enhanced, feeding back local information from the rover to increase its awareness of the scenario. Second, a computationally lightweight and collision-free motion planner for mobile manipulation is presented, which ensures the safety of the robot while performing the mobile manipulation scientific tasks. Third, an optimal-control-based motion planner for mobile manipulators is proposed, which can maximize the motion efficiency and comply with the system constraints, at the same time it keeps a low computational cost thanks to a series of warm start stages, which makes it suitable for space. Fourth, a model predictive controller for mobile manipulation is used to increase the safety of multirobot missions, in the particular use case of rappelling into lava tubes. All of these contributions have been thoroughly analyzed and validated by means of experimental campaigns, which include simulation, laboratory, and field tests with rover prototypes in analogue scenarios to Mars or the Moon.