Enabling dexterous physics-based manipulation via a learning framework for shared autonomy

Abstract

There is a growing need in the navy for technology that can assist with the everyday maintenance of maritime infrastructure. Several of these tasks involve complex interactions with tools and with the world, including turning valves, calibrating gauges, cleaning and operating switchboards, among others. The first thrust of the proposal is to enable physics-based reasoning for autonomous manipulation. This capability is critical to free robots out of the factory floors and into real-world environments. The PI and team will address key challenges in this domain by building tractable models for reaslistic physics, using these models for manipulation planning, addressing robustness under uncertainty in actuator and pose estimates as well as physics model parameters, and incorporating feedback, especially via proprioceptive and tactile sensing. The second thrust of our proposal is to develop a mathematical optimal control framework for shared autonomy, where an expert user acts as a teacher, instructing and collaborating with the robot to accomplish a task. They will address key challenges in this domain by developing a formalism for shard autonomy, building mathematical models for shared autonomy via optimal control theory, and generalizing shared autonomy across user interfaces.

Document Details

Document Type
DoD Grant Award
Publication Date
Jun 03, 2016
Source ID
N000141612084

Entities

People

  • Siddhartha Srinivasa

Organizations

  • Massachusetts Institute of Technology
  • Office of Naval Research
  • United States Navy

Tags

Fields of Study

  • Computer science

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Distributed Systems and Data Platform Development

Technology Areas

  • AI & ML
  • AI & ML - Autonomous Systems
  • AI & ML - DoD AI Strategy
  • Autonomy
  • Autonomy - Autonomous System Control