News & Events

Habilitation lecture by Jan Faigl

Jan Faigl's habilitation lecture 'Multi-Goal Path Planning in Mobile Robotic Tasks'

takes place on  December 10, 2014  at 13:00 in room  T2:D3-209

 

Summary:

Multi-Robot exploration is a probem to acquire information about unknown environment by means of autonomous navigation of a group of mobile robots that use their sensor system to perceive the environment.
The fundamental approach to address the exploration is based on an iterative determination where to navigate the robots next, which is decomposed into determination of the suitable goal candidates and selection of the next navigational goal from these candidates.
In a case of a multi-robot team, this problem also includes an efficient allocation of the goals among the team members to support cooperation and coordination of the robots' motion,  which can be formulated as the task-allocation problem.
Moreover, the model of the environment being explored is continuously updated during the exploration mission, therefore it is desirable to continuously utilize the newest information available and perform a frequent replanning of the robots' actions.

New information about the environment can be acquired if the robots travel towards the unexplored parts of the environment, and therefore, goal candidates can be located at the border of the already explored and not yet explored parts of the environment.
These locations are called frontiers and various frontier-based approaches have been proposed.
The simplest approach based on greedy selection of the closet frontier to the robots has been improved by utility based evaluation that combines expected information gain with the distance cost.
In addition, novel approaches to reduce the number of goal candidates have been developed that not only allow to consider more computational demanding assignment procedures, but they also improve the performance for simple greedy based assignments.

On the other hand, several task-allocation algorithms have been deployed in the context of the exploration mission that include standard algorithms from the operational research as well as distributed approaches allowing decision-making under limited communication between the exploring units.
However, there is still lack of evaluation methodology to compare different exploration strategies that is not limited to particular experimental setup and that can provide a more general conclusion about the expected performance of the exploration strategy in various exploration missions.