Fail It till You Make It: Error Expectation in Complex-Plan Execution for Service Robots.

Domestic service robots (DSR) are devices aimed to carry out daily household chores. Recently, the range and difficulty of the activities they can perform are reaching amazing performances, frequently resulting in complex plans with several steps requiring many skills where errors are more likely to occur. In this paper, we introduce the concept of error expectation in complex plans for domestic service robots and propose a classification of DSR’s tasks based on the abilities required for their execution. We also propose a recovery system where a type of feedback is chosen based on error expectation and DSR task type. We test our proposal in the context of an object manipulation task and discuss how error expectation contributes to a good feedback choice. An action involving several robot skills, namely navigation, human and object recognition, natural language processing, etc., was illustrated by the “take” action. The video showing the complete execution of a complex command by a robot using error recovery is presented.