A Humanoid Robot for Investigating Spatial Understanding in Human-Robot Collaboration
Abstract
This project investigates spatial comprehension and modeling in the context of human-robot interaction, with the goal of achieving safe, autonomous interactions with people on a wide range of physical tasks. Achieving true autonomy requires increasing the robot~s understanding of the dynamics of its world (physical understanding), as well as the actions of people (cognitiveunderstanding). Our system~s cognitive understanding arises from using a cognitive architecture as the reasoning and planning component. The system~s physical understanding stems from its mental model, which is a 3D virtual world that the architecture synchronizes with the environment in real time. This mental model represents changes in causation in the world and generates expectations about physical change in the world. The main scientific focus of thisproposal is the development of the 3D virtual world and its use in planning, learning and communication.The usual approach to the use of vision in robotics attempts to solve two problems: (a) Process visual data to extract all the objects and motions in the environment,(b) Identify the results from (a) that are important and relevant to the current task.Unfortunately, both of these steps are very expensive computationally. The first step requires processing an enormous amount of visual data, especially when the environment is very dynamic. The second step is a difficult data mining problem. Our approach to this complexity issue is to leverage goal-directed rendering: the robot first decides which aspects of its environment are relevant, based on its task goals. This information is used to focus the cameras on specific regions of the environment and extract only the information needed for the goals. Because the system ignores irrelevant information, it is fasterand less expensive than current approaches. Our current system runs on a laptop in real time. Our goal is to create a small, inexpensive 3D modeling system.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 30, 2019
- Source ID
- N000141912546
Entities
People
- David Benjamin
Organizations
- Office of Naval Research
- Pace University
- United States Navy