An Object Template Approach to Manipulation for Humanoid Avatar Robots for Rescue Tasks

Abstract

Nowadays, the first steps towards the use of remote mobile robots to perform rescue tasks in disaster environments have been made possible. However, these environments still present several challenges for robots, which open new possibilities for research and development. For example, fully autonomous robots are not yet suitable for such tasks with high degree of uncertainty, and pure teloperated robots require high expertise and high mental workload, as well as fast communication to be reliable. In this paper, we discuss a middle ground approach to manipulation, that leverages the strengths and abilities of a human supervisor and a semi-autonomous robot while at the same tackling their weaknesses. This approach is based on the object template concept, which provides an interaction method to rapidly communicate to a remote robot the physical and abstract information for manipulation of the objects of interest. This approach goes beyond current grasp-centered approaches by focusing on the affordance information of the objects and providing flexibility to solve manipulation tasks in versatile ways. Experimental evaluation of the approach is performed using two highly advanced humanoid robots.