IV

Mobile Robots in ROS

INTRODUCTION

Mobile-robot navigation can be subdivided roughly into path planning and driving. A path planner is responsible for proposing a collision-free, efficient and navigable path to be followed. A driving subsystem is responsible for controlling the robot to execute such a plan safely and precisely. Achieving good navigation requires development of multiple, inter-related subsystems.

Global path planning requires a model (map) of the environment, and thus mapmaking and representation are important topics.

Global path planning typically is performed with simplifying assumptions on dynamics, resulting in coarse plans. These plans are converted to efficient, achievable trajectories through trajectory planning.

Steering algorithms are used to achieve precise following of plans, but this requires knowledge of path-following errors to be corrected. Localization is thus a crucial module used to obtain robot coordinates from which one can compute path-following errors.

The subject of path planning branches into a variety of cases, including a priori specification of a path to be followed, autonomous path planning based on knowledge of a map, planning with uncertainty or partial map information, local path planning in unknown environments with global positioning available, or exploration behaviors without benefit of maps or global positioning information.

These topics are broad and are subjects of ongoing research. It is not presumed that this treatment will cover the field of mobile robotics. Rather, use of ROS for mobile platforms will be described with respect to illustrative algorithms. Further, the examples here will be limited to differential-drive ground vehicles, although the concepts are extensible to more complex systems, including holonomic vehicles, walking and crawling robots, aerial vehicles, and underwater vehicles.