Versatile Manipulation for Assistive Free-Flyers

The ability to manipulate the environment could add an important new dimension to the role of Assistive Free-Flyer (AFF) robots such as the NASA Astrobee. Co-located with the human crew on board the International Space Station (ISS), AFFs have shown potential to assist in areas like surveillance, inspection, mapping, etc. However, a platform without manipulation abilities is restricted to tasks where its main role is to deliver one or more sensors to a desired location. The capability to physically interact with the environment through manipulation could greatly increase AFF's reach, literally and figuratively. In this project, we proposed new approaches for the optimization and quantification of underactuated hand designs. These have allowed us to develop a new generation of robot hands that could enable versatile manipulation for AFFs in a small, lightweight package. Our new methods showed that we can optimize the design of a dexterous robot hand with many Degrees of Freedom (DoF) to achieve versatility in grasping a wide and user-specified set of objects even when powered by a single motor. In turn, this level of under actuation can allow the hand to meet the strict packaging constraints for fitting in small volumes such as the Astrobee payload bay. This was the first time that a design optimization algorithm was proposed that can explicitly optimize the transmission mechanism for such a hand so that it can achieve stable grasping of a user-defined set of objects of varying shapes. We demonstrated  this  novel conceptual  and  algorithmic  approach  by  building  a  set  of  progressively more advanced hands, the latest of which we tested and demonstrated on the Astrobee robot ground unit at NASA Ames. In the course of this project, this technology advanced from TRL 2 to TRL 6.Furthermore, the project achieved a number of additional results. Exploring the spectrum between highly underactuated and fully actuated sensorized hands, we showed that a fully actuated gripper can achieve versatile grasping based on proprioception alone, without any vision or tactile information. We also made significant steps towards a software framework for surface-to-orbit teleoperation of versatile manipulation using underactuated hands. Taken together, these advances build up towards the goal of achieving versatile intravehicular manipulation in orbit using passively compliant and underactuated robot hands.