| Autonomous robot systems require the recovery of robust spatial descriptions from sensory information and their efficient utilization in robot planning and control tasks. Traditional approaches to robot perception have emphasized the use of geometric sensor models and heuristic assumptions to constrain the sensor interpretation process, and the use of geometric world models as the basis for planning robotic tasks. These approaches, however, are of limited use in complex scenarios, such as those encountered by autonomous mobile robots operating in unknown and unstructured environments.;We then discuss the application of the Occupancy Grid framework to a variety of mobile robot perception, navigation, spatial reasoning and control tasks. These include range-based mapping, multi-sensor integration, path-planning and obstacle avoidance, handling of robot position uncertainty, incorporation of pre-complied maps, recovery of geometric representations, and other related problems. The experimental results show that the Occupancy Grid approach provides dense world models, is robust under sensor uncertainty and errors, and allows explicit handling of uncertainty. It supports the development of robust and agile sensor interpretation methods, incremental discovery procedures, and composition of information from multiple sources. Furthermore, the results illustrate that robotic tasks can be addressed through operations performed directly on the Occupancy Grid, and that these operations have strong parallels to operations performed in the image processing domain. We conclude with a discussion of the role of high-level and low-level models in robot perception.;In this thesis we introduce a new framework for spatial robot perception, real-world modelling, and navigation that uses a stochastic tesselated representation of spatial information called the Occupancy Grid. The Occupancy Grid is a multi-dimensional random field that maintains probabilistic estimates of the occupancy state of each cell in a spatial lattice. To recover a sensor-based map of the robot's environment, the Occupancy Grid is estimated using stochastic sensor models. A Bayesian estimation procedure allows the incremental updating of the Occupancy Grid, using readings taken from several sensors and over multiple points of view. Additional stochastic estimation methods provide mechanisms for composition of multiple maps, integration of information from different sensors, decision-making, and handling of robot and sensor position uncertainty. |
