Distributed Robotics

Distributed Manipulation

In robotics most manipulation algorithms are designed to execute in a single process on a single computer that takes input from all the sensors and controls all the effectors. These algorithms are sequential but mechanically-based single robot algorithm. While quite general in principle, these off-line algorithms are usually designed for robotics devices such as grippers or fingers attached to a traditional robot arm. To extend these results for distributed manipulation there are several challenges. (1) Autonomous mobile robots (mobots) have different sensing and control modalities. (2) Mobots are often better suited to on-line approaches and hence the algorithms must be adapted to rely less extensively on geometrical and dynamical models. (3) A host of difficulties arises when a task must be performed by a distributed team instead of a single agent ---and hence the algorithms must compensate by changing their communication, sensing, or knowledge requirements.

Projects: