Robot Behavioral Interaction on Rough Terrain

Abstract

This project studies the structure of a robotÕs dynamic interaction with challenging terrain and develops new methods for a robot to reliably operate in the presence of uncertainty. Outside of a lab or factory, the environment that a robot is operating in can not be controlled or perfectly modeled. As a result, the behavior of the overall system as the robot contacts the environment will always have some degree of underaction and undersensing. To improve the robustness of the behaviors on challenging terrain, this project will seek to both reduce the amount of uncertainty in the system as well as minimize the effect that any remaining uncertainty can have on the overall results. Especially for legged robots, changing contact conditions can lead to discontinuities in the behavior. A foot that misses a planned contact or a knee that unexpectedly hits a rock can easily lead to failure. The dynamics of such a system are hybrid in nature, combining a discrete contact mode and continuous configuration space. This project will study the underlying structure or topology of these hybrid dynamical systems to understand the nature of these critical contact events. This structure can then be used to generate more reliable robot control plans. The value of this research will be demonstrated in the context of a particularly challenging robot mobility task: climbing steep, rocky hills. Experiments in simulation and on robotic hardware will show reduced sensitivity to uncertainty in both the state and the dynamical model, as well as a reduced frequency of unexpected contact changes. The results of this project will be both an improved theoretical understanding of the fundamental structure of robot-terrain dynamics as well as more reliable operation for the application of rocky hill climbing robots.

Document Details

Document Type
DoD Grant Award
Publication Date
Feb 14, 2019
Source ID
W911NF1910080

Entities

People

  • Aaron M. Johnson

Organizations

  • Army Contracting Command
  • Massachusetts Institute of Technology
  • United States Army

Tags

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Robotics and Automation.
  • Systems Analysis and Design

Technology Areas

  • AI & ML
  • AI & ML - Autonomous Systems
  • AI & ML - Bayesian Inference
  • Autonomy
  • Autonomy - Autonomous System Control
  • Space
  • Space - Spacecraft Maneuvers