This paper introduces a novel and general real-time adaptive motion planning (RAMP) approach suitable for planning trajectories of high-DOF or redundant robots, such as mobile manipulators, in dynamic environments with moving obstacles of unknown trajectories. The RAMP approach enables simultaneous path and trajectory planning and simultaneous planning and execution of motion in real time. It facilitates real-time optimization of trajectories under various optimization criteria, such as minimizing energy and time and maximizing manipulability. It also accommodates partially specified task goals of robots easily. The approach exploits redundancy in redundant robots (such as locomotion versus manipulation in a mobile manipulator) through loose coupling of robot configuration variables to best achieve obstacle avoidance and optimization objectives. The RAMP approach has been implemented and tested in simulation over a diverse set of task environments, including environments with multiple mobile manipulators. The results (and also the accompanying video) show that the RAMP planner, with its high efficiency and flexibility, not only handles a single mobile manipulator well in dynamic environments with various obstacles of unknown motions in addition to static obstacles, but can also readily and effectively plan motions for each mobile manipulator in an environment shared by multiple mobile manipulators and other moving obstacles.

This paper considers the problem of planning motions for a team of mobile manipulators working in the same environment with a common task objective. It presents a distributed, real-time algorithm to plan motion trajectory for each team member that allows dynamic and spontaneous division of work among team members to meet the common task objective. A mobile manipulator has to perform its share of the task while avoiding other moving mobile manipulators in the team in addition to other obstacles in the environment. To each robot team member, none of the trajectories of the other team members or moving obstacles are known beforehand. The approach is implemented and tested in simulated task environments, which demonstrates its high effectiveness and efficiency.

This paper considers the problem of planning closed-chain motion for a pair of mobile manipulators to transport a common payload in a dynamically unknown environment (i.e., an environment with moving obstacles of unknown motion). We present a novel algorithm to plan the actions of the two robots in the team, one leader and one helper, in real-time to accomplish the task while avoiding other obstacles in the unknown dynamic environment. Our algorithm does not assign fixed roles to the two team members, but rather it dynamically decides who should lead based on circumstances to optimize the team’s performance. The planner is readily extensible to handle a team of more than two robots in tight collaboration. The approach is implemented and tested in simulated task environments, which demonstrate the planning algorithm’s effectiveness and efficiency.

This video demonstrates a real-time adaptive motion planner for a mobile manipulator to accomplish place-to-place tasks in a dynamic environment with obstacles of unknown motion. Paths and trajectories are planned simultaneously as the robot moves and globally subject to some optimization criteria based on evolutionary computation. The robot always follows the current best trajectory with respect to predictions of obstacle motions through sensing. At any time the robot may switch seamlessly to a better trajectory as the planner continues to improve or adapt trajectories to the changing environment. To minimize energy and time, the planned arm and base trajectories are loosely-coupled so that the arm may stop its motion (relative to base) for some period while the base moves, or vice versa, in handling obstacles.

While there has been a large body of literature addressing offline path planning for manipulators, there is relatively less study on real-time motion planning that occurs as a manipulator moves in an environment with unknown obstacles or unknown changes. This paper introduces a unified and general motion planning approach based on evolutionary computation that is suitable for both offline and real-lime adaptive motion planning for manipulators under various optimization criteria and manipulator constraints in environments with obstacles or changes not known a priori. The implementation and testing results demonstrate the effectiveness and efficiency of the approach.