Extended reliable robust motion planners

Abstract

A new method to plan guaranteed to be safe paths in an uncertain environment, with an uncertain initial and final configuration space, while avoiding static obstacles is presented. First, two improved versions of the previously proposed BoxRRT algorithm are presented: both with a better integration scheme and one of them with the control input selected according to a desired objective, and not randomly, as in the original formulation. Second, a new motion planner, called towards BoxRRT*, based on optimal Rapidly-exploring Random Trees algorithm and using interval analysis is introduced. Finally, each of the described algorithms are evaluated on a numerical example. Results show that our algorithms make it possible to find shorter reliable paths with less iterations.

Reliable Motion Plannning for a Mobile Robot

“…This means that while planning, uncertainties usually resulting from: approximate localization, imperfect embedded sensors or approximate models used to describe the behaviour of the mobile robot devices should be taken into account. Considering uncertainties in the navigation planning level has been considered using several representations such as set-membership ( [13], [15], [17]) or covariance matrices ( [6], [16], [3]). While the latter is able to find paths with a collision probability under a given threshold, set-membership approaches can guarantee safe trajectories under a bounded noise assumption.…”

“…Recently, some studies provided the guarantee of a safe path to imperfect proprioceptive sensors and/or localization information, while considering the uncertainties bounded with know bounds ( [14], [17]). Under the latter uncertainty representation, [15] and [17] introduced reliable and robust navigation planners algorithms based on RRT principles and solved using interval analysis tools ( [10], [4]).…”

“…For the navigation planning level will make use, firstly, of a classical and well known planning algorithm (RRT) provided by the the open motion planning library (OMPL [19]). Secondly, a planning algorithm which can guarantee safe paths in an uncertain configuration space, where all approximate initial and final mobile robot localisation are bounded with known bounds, e.g., BoxRRT [17], will be considered. As in this paper the main focus is at the navigation planning algorithm, as stated previously, the navigation control level, i.e, the low-level one of the autonomous navigation architecture, will be based on PID controller along with the go-to-goal strategy, which can easily be formulated or found in the literature [18].…”

“…If it can be proved that the edge (typically a straight lines) between s near and s new is a collision free path, then s new is added to G as a new vertex G.addvertex procedure and connected to its parent s near though the G.add-edge procedure. Several versions of BoxRRT can be found in [15], [14], [17] for the case where the uncertainties related to the configuration space are considered on the final and the initial states. In this paper, we have interest in using as the navigation planner level the version of BoxRRT proposed in one of our recent study [17], which also presents the differences between the robust motion planners.…”

Reliable Navigation Planning Implementation on a Two-Wheeled Mobile Robot

“…A challenge for the autonomous navigation's planner level is related to the guarantee of the system's safety. To this end, in this study, we use a previously proposed reliable motion planner algorithm [3] based on Rapidly-exploring Random Trees (RRT) principles [1], which covers the whole configuration space and easily integrates complex robot models, and solved in an interval analysis [2] framework. More precisely, we consider while planning, the uncertainties from a visual localization provided by camera, bounded with know bounds.…”