Reactivity

Abstract

Robots have particular roles in their environments [130]. The current and future roles of robots include nuclear waste removal, office and hospital courier service, carpet cleaning in buildings, household tasks, spraypainting in automatic factories, and planetary exploration. Fulfilling such different roles requires specialization in robot hardware: a robot acting in rough terrain needs legs whereas wheels suffice for an office robot. Like the hardware, the control software should also — to some degree — be tailored to specific roles and environments. A planetary-exploration robot should choose its actions differently than an office courier or a manufacturing robot. The planetary explorer has a fixed mission in which the robot’s safety is one of the highest priorities but it does not need to execute its activities very efficiently. An office robot, on the other hand, must accomplish diverse jobs in a more fast-paced environment in which it is not the only agent. It is unrealistic to expect one kind of robot controller to work well in such different jobs and environments. Moreover, many difficult robot control and perception problems can be solved more easily in specific habitats [98, 97], especially when the habitats are designed to facilitate jobs [3]. To summarize, the desirable features of a robot controller are to a large degree determined by the capabilities of the robot and the features of its jobs and environment.