Technical Report: Reactive Planning for Mobile Manipulation Tasks in Unexplored Semantic Environments

Complex manipulation tasks, such as rearrangement planning of numerous\nobjects, are combinatorially hard problems. Existing algorithms either do not\nscale well or assume a great deal of prior knowledge about the environment, and\nfew offer any rigorous guarantees. In this paper, we propose a novel hybrid\ncontrol architecture for achieving such tasks with mobile manipulators. On the\ndiscrete side, we enrich a temporal logic specification with mobile\nmanipulation primitives such as moving to a point, and grasping or moving an\nobject. Such specifications are translated to an automaton representation,\nwhich orchestrates the physical grounding of the task to mobility or\nmanipulation controllers. The grounding from the discrete to the continuous\nreactive controller is online and can respond to the discovery of unknown\nobstacles or decide to push out of the way movable objects that prohibit task\naccomplishment. Despite the problem complexity, we prove that, under specific\nconditions, our architecture enjoys provable completeness on the discrete side,\nprovable termination on the continuous side, and avoids all obstacles in the\nenvironment. Simulations illustrate the efficiency of our architecture that can\nhandle tasks of increased complexity while also responding to unknown obstacles\nor unanticipated adverse configurations.\n