Autonomous Mobile Robots and Multi-Robot Systems

7Motion in an Unknown Environment

Eugene Kagan

This chapter considers the methods of probabilistic motion planning in the unknown environment widely known as probabilistic robotics. It introduces the concept of belief space, considers basic estimation and prediction methods and additional methods of the environment mapping. In addition, the chapter presents the simplest learning methods and their implementations for the robots' control.

7.1 Probabilistic Map‐Based Localization

In contrast to the fixed‐base arm manipulators, which act in artificial well‐structured environments and because of practically unlimited power supply are equipped with robust heavy‐duty gear, mobile robots usually act in real unstructured worlds and have relatively simple economical and consequently inaccurate equipment that leads to errors and uncertainties in the mobile robots behavior. In the widely accepted systematization, the sources of such uncertainties are listed as follows (Thrun, Burgard, and Fox 2005):

1. Environments. The real‐world environments, in which the mobile robots act, are highly unpredictable and usually change in time.
2. Sensors. The sensors are limited in their perception abilities, and the obtained measurements are perturbed both by the environmental noise and by the internal errors.
3. Actuators. The actuators usually driven by motors are not accurate and are perturbed by external noise. Additional errors are introduced by the control noise.
4. Models. Any model provides noncomplete ...