Teaching Task Flow Through Dialog and Observation

摘要: In order for robots to act as valuable assistants non-expert users, they need be able learn new abilities and do so through natural methods of communication. Furthermore, it is often desirable that tasks learned quickly without having provide multiple demonstrations. Training should also conducted the user has a clear understanding manner in which environmental features affect behavior activity, execution predictable. We present an interactive framework teaching robot flow activity composed elements from set primitive behaviors previously trained activities. Conditional branching looping, order-independent execution, contingency (or interrupt) actions can all captured by our structures. Additional convenience functionality aid training process provided. By providing method communicating production rules analogous rigid programming structures, easily. demonstrate task procedure on CMAssist mobile robot.