Learning Object Affordances by Leveraging the Combination of Human-Guidance and Self-Exploration

摘要: Our work focuses on robots to be deployed in human environments. These robots, which will need specialized object manipulation skills, should leverage end-users efficiently learn the affordances of objects their environment. This approach is promising because people naturally focus showing salient aspects [1]. We replicate prior results and build them create a combination self supervised learning. present experimental with robot learning 5 4 using 1219 interactions. compare three conditions: (1) through self-exploration, (2) from examples provided by 10 naive users, (3) self-exploration biased user input. characterize benefits affordance show that combined most efficient successful.