On-line deadlock-free path-planning algorithms by means of a sensor-feedback tracing

Abstract

In the sensor-based path-planning, a mobile robot always finds a deadlock-free path towards the goal if such the path exists in an uncertain world, otherwise it automatically sees that there is no deadlock-free path in the world. All the previous sensor-based path-planning algorithms keep two terminations as their deadlock-free characteristic under the following assumptions: 1) a robot faithfully traces an uncertain obstacle at a constant distance; and 2) a robot exactly recognizes its position. However, in general, a mobile robot with the non-holonomic system cannot faithfully trace an uncertain obstacle. Thus in this paper, we discuss deadlock-free characteristics of those algorithms after removing the first assumption in the presence of sensing and control errors. Then we propose some robust versions from the previous algorithms and also give proofs of their deadlock-free characteristics by a non-contact obstacle tracing with some errors.