| DENEB Robotics | ||||
| mazer@imag.fr |
| Issues to Robot Motion Planning |
|
Part 1: Overview Deneb Robotics' simulation an telerobotic software is widely used is aerospace, automotive, defense, environmental, medical, nuclear and research communities. However the use by custumers of Deneb automatic planners are still quite limited. Indeed the expectations of industrial users towards such automatic planners are generally very high and difficult to met. Most of the time the solutions obtained by experienced operator with sophisticated simulation tools challenge by far the solutions obtained by automatic methods both in term of speed an quality. Using Deneb simulation tools operators are capable to quickly devise reasonable trajectory which will be difficult to obtain from automatic planners. In some occasion, the contrary may also be true : an automatic planner may solve a problem taken as impossible for a human to solve. We will present such an example and comment on the requirements for an automatic planner to gain support among end-users. Part 2: Key research issues Fundamental studies are needed to better understand the tremendous computational explosion occuring for just some examples given a planner. Is there a way to avoid this kind of behavior or is it foolish pretending to attack NP problems for industrial applications ? Answering this question will give some hints about possible future research directions. People believing the first hypothesis may define slightly modified version of the planning problem to make it less difficult. Or they may look at the percolation theory to find when things start to be really difficult, or they may consider playing "path planning" : I give you a planner give me a deceptive example : if you can't I win ! Or they may consider a probabilistic version of this game and make it more interesting by playing for stakes. People believing the second hypothesis will probably be more interested in developing strategies helping the operator to solve the problem. They could develop sophisticated man machine interface, powerful local planner or new type of visualization which will help the operator to solve his problem. They will concentrate on the geometrical engines to perform these operations fast. Part 3: Scenarios and benchmarks Deneb may provide difficult industrial examples on its platform and may consider to adapt and open it's geometric engine to suite a particular planner needs. This will allow several research group to compare their approaches on industrial benchmarks. |