| dc.description.abstract | Multi-robot systems have many uses such as cleaning, exploration, search and rescue. These
robots operate under constraints such as communication, battery etc. In this thesis, we provide a
method by which the robots can hand-off their current task to a new robot so that the given task
can be continued without interruption. It is assumed that the task can be handed off to any other
robot without losing the progress on the task. In the task hand-off framework, the robots complete
as much of the task as possible before trying to replenish their resources (e.g., refuel). The robots
must also make sure that the task is handed over to another robot before they go back to refuel.
We demonstrate the task hand-off framework in the context of a battery constraint. The robots
hand-off their current task once they are low on battery. The robots are divided into helpers and
workers. The workers are the ones that perform the given task while the helpers wait at charging
locations. Once a worker determines it is running out of battery it calls for help and switches
behaviors with a helper. The new worker then takes over the task. This framework allows a user
to model robot teams performing common robotic tasks such as exploration, coverage or any other
task where the task can be easily handed-off without losing any progress on the task.
We also present a simple priority based inter-robot contention resolution algorithm using motion
replanning to avoid inter-robot collisions. Each robot is assigned a priority. Whenever the
robots are close to each other, the lower priority robots halt and the highest priority robot replans a
path around the robots by considering them as additional robots.
We demonstrate the task hand-off framework approach using a physics based simulator that is
built on top of a physics engine and also using physical hardware. The physical hardware consists
of multiple iRobot Create robots with an onboard ASUS Netbook. We provide results from room
407 of the Harvey Bum Bright Building at Texas A&M University. We show that the tasks get
completed faster with task hand-off than when task hand-off was not allowed. | en |
