Error Recovery Patterns Focusing on the Revival Process from Failures in Manipulation Tasks

Abstract

In recent years, robots have played an important role in various places, including factories of the manufacturing industry as well as homes where people live. The number of robotic tasks with a high degree of difficulty is increasing because they are required to perform various types of tasks, and failures are likely to occur. Therefore, there is a growing demand for error recovery techniques. We propose an error recovery method that considers task stratification and error classification. This allows various error-recovery paths to be derived for a single error. In this study, recovery paths were classified into several patterns to facilitate selection of the optimal path.