Navigating mobile robots

Figure 1.6: (a) Several mobile robots attempt to successfully navigate in an indoor environment while avoiding collisions with the walls and each other. (b) Imagine using a lantern to search a cave for missing people.

Figure 1.7: A mobile robot can reliably construct a good map of its environment (here, the Intel Research Lab) while simultaneously localizing itself. This is accomplished using laser scanning sensors and performing efficient Bayesian computations on the information space [351].

A more common task for mobile robots is to request them to navigate in an indoor environment, as shown in Figure 1.6a. A robot might be asked to perform tasks such as building a map of the environment, determining its precise location within a map, or arriving at a particular place. Acquiring and manipulating information from sensors is quite challenging and is covered in Chapters 11 and 12. Most robots operate in spite of large uncertainties. At one extreme, it may appear that having many sensors is beneficial because it could allow precise estimation of the environment and the robot position and orientation. This is the premise of many existing systems, as shown for the robot system in Figure 1.7, which constructs a map of its environment. It may alternatively be preferable to develop low-cost and reliable robots that achieve specific tasks with little or no sensing. These trade-offs are carefully considered in Chapters 11 and 12. Planning under uncertainty is the focus of Part III.

If there are multiple robots, then many additional issues arise. How can the robots communicate? How can their information be integrated? Should their coordination be centralized or distributed? How can collisions between them be avoided? Do they each achieve independent tasks, or are they required to collaborate in some way? If they are competing in some way, then concepts from game theory may apply. Therefore, some game theory appears in Sections 9.3, 9.4, 10.5, 11.7, and 13.5.