The paradox of service robots-how passers-by can contribute in solving non-deterministic exceptional conditions encountered by service robots

Intelligent, autonomous service robots face unique challenges. Unlike autonomous guided vehicles used in manufacturing facilities, service robots cannot rely on clearly laid assumptions about the environment, as they share the work-space-which has not been designed around them-with human beings. At the same time, they cannot be treated as entertainment robots, since they are designed to accomplish a given set of tasks and their fate is measured by success-rate and not by their cuteness. Service robots need to perform reliably even during exceptional situations. The need for reliability strives for predictability, while the presence of a shared environment yields nondeterministic situations. The clash may be described as the paradox of mobile robotics, where reliability and safety are undermined by the same goals we want to reach. Based on the experience collected with the mobile mail distribution system MoPS, this paper describes a novel approach that is aimed at resolving the large number of error situations that can be addressed by untrained individuals, by allowing the robot to interact with passersby, if an exceptional situation is detected.