A Mission Planning System for Multiple Mobile Robotics in Unknown, Unstructured, and Changing Environments

Abstract

Research in autonomous mobile robots has reached a level of maturity where robotic systems can be expected to efficiently perform complex missions involving real-world tasks, particularly those which are expensive or risk-intensive, efficient teams of autonomous cooperative mobile robots could provide a valuable alternative to current solutions. Through the distribution of computation, perception, and action, a cooperative robot team is more capable than the sum of its parts, as this team exhibits increased reliability and the ability to complete physically distributed tasks. For mobile robots to be effective in real-world applications, more than one robot must be able to sagely share a potentially unknown workspace. Complicated missions with interdependencies between these robots must be feasible. Finally, robotics systems must accommodate an operational environment which is not necessarily static, certain, or known in advance. Many tasks which are likely candidates for robotic automation (such as hazardous waste site remediation, planetary exploration, materials handling and military reconnaissance), require a robot team to perform an essentially mobile mission which involves robots moving between significant locations. It is important that these missions be completed efficiently, appropriately minimizing the cost of the task. The similarities among these tasks indicate that a single general system could support coordinated mission execution for many scenarios. To this end, GRAMMPS (a General Robotic Autonomous Mobile Mission Planning System) has been developed. GRAMMPS supports the optimization of real-world missions involving multiple robots and multiple concurrent goals. The largest component of GRAMMPS is its central planner, which continuously optimizes the execution of a multi-robot mission as information about the world is acquired. GRAMMPS distributes its computation, gracefully degrades from optimal performance when presented with computationally intractable missions, and performs efficient replanning in an unknown, unstructured, and changing environment. This system has been demonstrated on two autonomous outdoor mobile robots and extensively validated in simulation.