0
$\begingroup$

What are robot task planners? And how are they implemented? I don't think they are just FSM or behavior tree right? What are the popular and robust task planning algorithms? And what are the state of the art algorithm?

I'm guessing task planning does priority assessment then choose to do certain task or not, for example, delivery robot that choose the optimal orders to drop-off.

$\endgroup$
1
1
$\begingroup$

Task and motion planning describes different levels of abstraction in a domain. Motion planning is about lowlevel actions for example move the car forward for 1 meter, while task planning is about high level actions for example to overtake another car. Task planning is usually treated under the umbrella term PDDL which is a dedicated planning language for formalizing abstract tasks. Or to be more specific: PDDL creates a simulation in which actions can be executed. The reason why task and motion planning are analyzed separate is to reduce the state space. So called hierarchical planners are providing a much better sampling efficiency than a normal motion planner.

For example, a car should overtake another car. On the first look this can be planned with moving the car forward by a half meter. And the complete maneuver consists of many repetitions of this action. The result will be a large state space which produces a huge load for the planner. The better idea is to see the sequence as a separate high-level task and not try to use a low level planner for solving it.

All organizations which are focused on plan-based robotic control have experience with PDDL. Either with the original specification or with it's derivative like OWLS-Xplan, OPL and Golog.

$\endgroup$
1
  • $\begingroup$ Do you know any organization that actually use PDDL? $\endgroup$ – drerD Mar 4 '19 at 7:44

Not the answer you're looking for? Browse other questions tagged or ask your own question.