Scene Representation Updating via Robotic Tactile Property Sensing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing scene representation models in computer vision and robotics are limited in accurately reflecting real-world scenes as they only partially describe the nature of objects, relying on geometry, color, and user-assigned semantic classes, which are insufficient for autonomous task execution by robots.
Innovation Solution
A computer-implemented method for updating a scene representation model by obtaining a value of a physical property of an object through physical contact with a robot, and then updating the model based on this obtained value, allowing for more complete representation of objects and improved autonomous task execution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If scene representation models use only geometry, color, and user-assigned semantic classes, then the model construction is simple and fast, but the accuracy and completeness of object representation is insufficient
Solution Approach 1:
The patent replaces visual sensing (optical system) with tactile sensing (mechanical contact system). The robot uses physical contact through its end effector to measure physical properties of objects, substituting the optical field-based representation with mechanics-based measurement. This allows acquisition of accurate physical property data (mass, stiffness, friction) that cannot be obtained through visual means alone, thereby improving representation accuracy while maintaining relatively simple implementation through contact-based sensing.
2Extent of automation
If user-assigned semantic class labels are used, then the scene representation can be generated with existing methods, but autonomous task execution is limited due to requiring user input
Solution Approach 1:
The robot performs self-service by autonomously exploring the scene and physically contacting objects to discover their physical properties without user intervention. The system independently generates the scene representation model by having the robot navigate, select objects, make contact, and measure properties automatically. This eliminates the need for user-assigned semantic labels while achieving complete object characterization through autonomous tactile exploration.
3Loss of information
If physical contact measurement is implemented, then complete physical property data can be obtained autonomously, but the device complexity and measurement time increase
Solution Approach 1:
The robot applies partial contact actions (light touches, probes) rather than attempting to fully manipulate or move each object. By using minimal contact forces and simple measurement interactions, the system efficiently extracts physical property information without excessive time investment. The contact measurements are performed selectively on key properties rather than exhaustive characterization, balancing information completeness with time efficiency.
Data Source
AI summary
A computer implemented method for updating a scene representation model is disclosed. The method comprises obtaining a scene representation model representing a scene having one or more objects, the scene representation model being configured to predict a value of a physical property of one or more of the objects; obtaining a value of the physical property of at least one of the objects, the obtained value being derived from a physical contact of a robot with the at least one object; and updating the scene representation model based on the obtained value. An apparatus is also disclosed.