Robot Sensor Repositioning for Close-Up Object Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Robotic systems face challenges in accurately interacting with objects in their environment due to limited sensor data from remote sensors, particularly when tasks require close-up information, such as identifying contents within opaque containers or determining user-preferred manipulation methods.
Innovation Solution
Equipping robotic gripping devices with positionable non-contact sensors, like infrared microcameras, that can be positioned at predetermined poses relative to objects based on object type, allowing for more accurate sensor data collection and input into machine-learned classifiers for property determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If remote sensors are used for object detection, then the robot can detect objects from a distance, but the sensor data lacks the detailed close-up information needed for accurate classification and property determination
Solution Approach 1:
The sensor system is made dynamic by enabling the sensor to move between different positions (remote and close-up) based on the operational requirements. The sensor can be repositioned along the robotic arm to achieve the predetermined pose relative to the object, allowing the system to adapt between distance detection and detailed classification tasks.
Solution Approach 2:
The robotic arm serves as an intermediary mechanism that positions the sensor at the optimal location relative to the object. By using the robotic arm as a mediator, the system can achieve precise sensor positioning without direct manual intervention, resolving the contradiction between measurement precision and ease of operation.
2Loss of information
If multiple sensors are positioned at different locations, then comprehensive object data can be collected, but the system complexity and cost increase
Solution Approach 1:
A single sensor is designed to perform multiple functions by being repositioned to different locations. The same sensor can be used for both remote object detection and close-up classification tasks, eliminating the need for multiple dedicated sensors and thereby reducing system complexity while maintaining information completeness.
Solution Approach 2:
Instead of having multiple static sensors at fixed locations, the system uses one dynamic sensor that can move to different positions. This dynamic approach allows the single sensor to capture comprehensive object information that would otherwise require multiple sensors, reducing system complexity while preventing information loss.
3Measurement precision
If the sensor is positioned close to the object for detailed classification, then accurate property determination is achieved, but the robot's ability to detect objects from a distance is reduced
Solution Approach 1:
The sensor positioning is made dynamic, allowing the system to switch between remote detection mode and close-up classification mode as needed. The sensor can be quickly repositioned along the robotic arm to achieve the predetermined pose for detailed classification after initial object detection, thereby maintaining both detection capability and classification accuracy without significant time loss.
Solution Approach 2:
The system performs preliminary object detection from a distance using the remote sensor position to identify potential objects of interest. Once an object is identified, the sensor is then repositioned to the predetermined close-up pose for detailed classification. This preliminary action approach ensures that close-up positioning is only performed when necessary, minimizing time loss while maintaining high classification accuracy.
Data Source
AI summary
In one embodiment, a method includes receiving, from a first sensor on a robot, first sensor data indicative of an environment of the robot. The method also includes identifying, based on the first sensor data, an object of an object type in the environment of the robot, where the object type is associated with a classifier that takes sensor data from a predetermined pose relative to the object as input. The method further includes causing the robot to position a second sensor on the robot at the predetermined pose relative to the object. The method additionally includes receiving, from the second sensor, second sensor data indicative of the object while the second sensor is positioned at the predetermined pose relative to the object. The method further includes determining, by inputting the second sensor data into the classifier, a property of the object.