Robot Controller Holding Manner Estimation for Unregistered Objects
Solution Overview
Problem
Existing robot control systems face challenges in determining a uniform holding position for objects, leading to reduced holding feasibility and potential failure in secure or user-specific holding operations.
Innovation Solution
A robot control device equipped with a controller that estimates the holding manner of an object using a database and an inference model, allowing for the acquisition of recognition information and determining the appropriate holding position based on this information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the robot control system uses only a database with pre-registered object information, then the holding position determination is fast and reliable for known objects, but the system cannot handle new or unregistered objects
Solution Approach 1:
The patent introduces an inference model as an intermediary component between the database and the holding position determination process. When an object is not found in the database, the inference model acts as a mediator to estimate the holding position based on extracted features, thereby enabling the system to handle unregistered objects without requiring complete reconfiguration
Solution Approach 2:
The patent segments the holding position determination process into two distinct pathways: one for registered objects (database lookup) and one for unregistered objects (inference model). This segmentation allows the system to optimize each pathway independently, maintaining speed for known objects while enabling adaptability for new objects
2Adaptability or versatility
If the robot control system uses an inference model to estimate holding manner, then the system can handle diverse and unregistered objects, but the estimation accuracy may be reduced compared to database lookup
Solution Approach 1:
The patent performs preliminary actions by extracting features from object images and comparing them against the database before invoking the inference model. This preliminary check ensures that the inference model is only used when necessary, thereby maintaining high accuracy for registered objects while preserving adaptability for unregistered ones
Solution Approach 2:
The system incorporates feedback mechanisms where the results from inference model estimations can be used to update and refine the database over time. This feedback loop continuously improves the accuracy of both pathways, ensuring that the inference model becomes more reliable as the database grows
3Measurement precision
If the system acquires recognition information from multiple viewpoints, then the holding position determination becomes more accurate, but the information acquisition time and complexity increase
Solution Approach 1:
The patent applies partial action by acquiring recognition information from a limited set of key viewpoints rather than all possible angles. This selective approach captures sufficient geometric features for accurate holding position determination while significantly reducing the time and complexity compared to comprehensive multi-view acquisition
Data Source
AI summary
A robot control device includes a controller. The controller is configured to be capable of estimating a holding manner of a holding target using each of a database and an inference model, and to control a robot based on the estimated holding manner. The database contains reference information including object information of multiple objects and holding manner information of the multiple objects. The inference model is capable of estimating a holding manner of an object. The controller acquires recognition information of a holding target. When the controller determines that the holding manner of the holding target cannot be estimated from the database based on the recognition information, the controller estimates the holding manner using the inference model.