Search

This paper is focused on probabilistic occupancy grid mapping and motion planning such that a robot may build a map and explore a target area autonomously in real time. The desired path of the robot is developed in an optimal fashion to maximize the information gain from the sensor measurements on its path, thereby increasing the accuracy and efficiency of mapping, while explicitly considering the sensor limitations such as the maximum sensing range and viewing angle. Most current exploration techniques require frequent human intervention, often developed for omnidirectional sensors with infinite range. The proposed research is based on realistic assumptions on sensor capabilities. The unique contribution is that the mapping and autonomous exploration techniques are systematically developed in a rigorous, probabilistic formulation. The mapping approach exploits the probabilistic properties of the sensor and map explicitly, and the autonomous exploration is designed to maximize the expected map information gain, thereby improving the efficiency of the mapping procedure and the quality of the map substantially. The efficacy of the proposed optimal approach is illustrated by both numerical simulations and experimental results.

Two variations of the joint probabilistic data association filter (JPDAF) are derived and simulated in various cases in this paper. First, an analytic solution for an optimal gain that minimizes posterior estimate uncertainty is derived, referred to as the minimum uncertainty JPDAF (M-JPDAF). Second, the coalescence-avoiding JPDAF (C-JPDAF) is derived, which removes coalescence by minimizing a weighted sum of the posterior uncertainty and a measure of similarity between estimated probability densities. Both novel algorithms are tested in much further depth than any prior work to show how the algorithms perform in various scenarios. In particular, the M-JPDAF more accurately tracks objects than the conventional JPDAF in all simulated cases. When coalescence degrades the estimates at too great of a level, and the C-JPDAF is often superior at removing coalescence when its parameters are properly tuned.

Occupancy grid maps are spatial representations of environments, where the space of interest is decomposed into a number of cells that are considered either occupied or free. This paper focuses on occupancy grid mapping, which is to estimate the probability of occupancy for each cell based on range measurements from a known location. For a given probabilistic model of a range sensor, we propose a computationally efficient method to obtain an exact inverse sensor model, and it is utilized to construct a probabilistic mapping algorithm according to the Bayesian framework. Compared with the existing occupancy grid mapping techniques that rely on approximate, heuristic inverse sensor models, the proposed approach yields substantially more accurate maps for the same set of measurements. These are illustrated by numerical examples and experiments.