EURASIP Journal on Advances in Signal Processing - Special issue on signal processing advances in robots and autonomy
Rigidity guided localisation for mobile robotic sensor networks
International Journal of Ad Hoc and Ubiquitous Computing
| Hi-index | 0.00 |
A fundamental capability for autonomous robot operations is localization—that is, the ability of a robot to estimate its position in the environment. It is a base level capability enabling numerous other technologies including mapping, manipulation, and target tracking. Within this realm of research, there is a more recent and narrower focus on Cooperative Localization (CL) for robot teams. In this paradigm, groups of robots combine sensor measurements to improve localization performance. The focus of this dissertation is applying sensor fusion techniques to the CL problem. We are particularly interested in employing teams of robots in target tracking roles. This has motivated our own solution to CL which is capable of solving the Simultaneous Localization and Target Tracking problem. Under this paradigm, targets are merely viewed as “passive” team members that must be localized. The benefit of such an approach is that although the targets do not contribute measurements, they still contribute constraints to the localization process. We assume an unknown-but-bounded model for sensor noise, whereby bearing and range measurements are modeled as linear constraints on the configuration space of the robot team. Merging these constraints induces a convex polytope representing the set of all configurations consistent with sensor measurements. Estimates for the uncertainty in the absolute position of a single robot or the relative positions of two or more nodes can then be obtained by projecting this polytope onto appropriate subspaces of the configuration space. While recovering the exact projection can require exponential time, we propose a novel method for approximating these projections using linear programming techniques. We then further extend current localization methods to the problem of active target tracking. Recall that in the CL model, pose estimates are formed by combining information from multiple distributed sensors. This capability invites the following question: given that the robot platforms are mobile, how should they be deployed in order to maximize the quality of the estimates returned by the team? We present a generic theoretical framework, and practical computational approaches for tackling this problem.