Robot Operation Control Selection for Situation-Specific Interaction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional robot control systems fail to manage and modify operations effectively across varying environments and situations, particularly struggling with physical interactions and type-specific operations, leading to potential operation failures when modifications are made without considering their impact on other situations.
Innovation Solution
A robot control device utilizing a machine learning model that receives situation information about physical characteristics, status, and relationships to output success or failure values, enabling an operation control selector to identify and execute appropriate control routines for achieving operation objectives through sensor interaction with objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a machine learning model is used to implement flexible control of a target device in a wide range of situations, then adaptability is improved, but it becomes difficult to manage, evaluate, verify, and modify operation controls for each individual situation
Solution Approach 1:
The patent segments the operation control system into multiple individual operation controls, each associated with specific situation information. The machine learning model evaluates each operation control separately based on situation inputs, allowing independent management and modification of each control without affecting others. This segmentation resolves the contradiction by enabling flexible control through the model while maintaining simplicity in managing individual operation controls.
2Reliability
If operation control is modified to correct an operation error in a certain situation, then reliability is improved in that situation, but the modification may affect operation success in other situations
Solution Approach 1:
The patent applies local quality by associating each operation control with specific situation information characteristics. When modifying an operation control to correct an error in a particular situation, the modification is localized to that specific operation control-situation pairing. The machine learning model evaluates each operation control independently based on its associated situation information, ensuring that modifications to one operation control do not affect others, thus maintaining both reliability improvement and adaptability.
3Object-affected harmful factors
If conventional systems determine object presence and perform operations only in areas with no objects, then collision avoidance is improved, but the system cannot perform physical interaction operations such as pushing objects aside
Solution Approach 1:
The patent changes the parameter of situation information from simple object presence/absence to comprehensive characteristics including object type, position, posture, and physical properties. The machine learning model uses these detailed parameters to determine whether physical interaction operations are appropriate. This allows the system to distinguish between situations requiring collision avoidance and those requiring physical interactions like pushing objects, thereby improving both safety and operational versatility.
Data Source
AI summary
A robot control device is configured to control a robot and includes: a machine learning (ML) model for describing ranges of environments, or in other words, situations in each of which control for a control routine (operation control) executing an operation to achieve an operation objective is achievable; an operation control selector configured to select, based on an output value from the ML model, the operation control which is appropriate for the present situation; and an operation control executor configured to execute the operation control which has been selected. The operation control is a control routine for a robot to achieve an operation objective by the robot sensing a first object with interaction with a second object.