NRI: Towards Dexterous Micromanipulation and Assembly

Purdue University, West Lafayette IN

Investigators

Abstract

Robots that people are familiar with are large, roughly, human-sized. The theory and design tools for developing such macro-scale robots is well developed. By contrast, the theory and design tools applicable to tiny, or micro-scale, robots is nearly non-existent. The goal of this project is to enable the transition of robot manipulation technology from macro-scale robots to micro-scale robots. The expected project outcomes are: i.) controlled and predictable environments for micro-robotic manipulation and assembly; ii.) a new class of 3D vision-based micro-force sensors and a 3D multi-resolution vision system; and iii.) the identification of dexterous micro-manipulation primitives via human micro-teleoperation with new novel haptic probes. These results will enable the assembly of micro-scale systems that are currently not possible. Such systems are applicable across a wide range of domains, such as cm-to-mm scale robots, micro-sensors, steerable catheters, micro-fluidic, and energy harvesting devices. At the micro-scale, surface forces dominate the interactions causing unpredictable forces. This project aims to lay the foundations for tools, such as simulators, motion planners, controllers, etc., to be developed for this unique micro-scale environment. The research approach is to reduce the uncertainty in forces present in the micro-world to enable dexterous micro-manipulation and assembly. A new class of manipulation substrates, fixtures, micro-parts, and manipulation tools to control and overcome the levels of adhesion forces present in the micro-world will be created. To enable force control, 3D vision-based micro-force sensing probes along with a multi-resolution 3D vision system to detect the micro-forces in real-time will be developed. Dexterous micro-manipulation primitives will be identified from a human tele-operating a multi-probe micro-manipulation system with micro-force feedback in an augmented reality system. How much and what types of micro-force feedback information is needed for the human to perform different tasks will be studied. These motion primitives together with physics-based simulators can reduce uncertainty in motion planners. The insights gained here will dictate the force-control algorithms implemented in future automated systems.