Robot Control With Flexible Feedback for High-Speed Assembly
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for robot control in assembly operations, such as product assembly in factories, face challenges in achieving high-speed contact operations with precision, especially when dealing with multiple parts and small components, due to limitations in positioning accuracy, force control, and the need for complex setup and calibration.
Innovation Solution
A control apparatus for robots incorporating a physically flexible portion and a learning model based on machine learning to determine optimal actions for assembly tasks, using state observation data to adjust the robot's movements and maintain contact with the environment, allowing for high-speed operations without requiring complex force control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high precision machines and mechanisms are used to reduce uncertainty in position and orientation, then positioning accuracy is improved, but device complexity and design adaptability deteriorate
Solution Approach 1:
The patent replaces complex mechanical positioning mechanisms with a learning model-based control system. Instead of using high-precision machines and dedicated jigs for each assembly target, the system uses machine learning to predict and compensate for positioning errors, thereby achieving high accuracy without complex mechanical structures.
Solution Approach 2:
The patent changes the control parameters from fixed mechanical positioning to dynamic learning-based adjustment. The learning model continuously adapts positioning parameters based on observed errors and feedback, enabling high precision without requiring complex mechanical mechanisms for each specific task.
2Adaptability or versatility
If visual sensors are used to estimate position and orientation, then adaptability is improved, but measurement precision deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where the learning model continuously receives observation data about the robot's state and the environment, processes this information, and adjusts control actions accordingly. This feedback loop enables the system to maintain high measurement precision while remaining adaptable to different tasks, as the learning model learns from accumulated experience rather than relying solely on sensor accuracy.
3Measurement precision
If force control is used to control application of force, then control precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent replaces complex force control mechanisms with a learning model that predicts appropriate force application based on observed states. Instead of using high-speed force control systems that require complex hardware and calibration, the learning model learns optimal force application strategies from experience, achieving precise force control without the associated complexity and cost.
4Stability of the object's composition
If passive operation mechanisms are used to absorb errors, then positioning tolerance is improved, but manufacturing precision requirements increase
Solution Approach 1:
The patent uses feedback through the learning model to compensate for positioning errors without requiring passive error-absorbing mechanisms. The learning model observes positioning deviations and learns to correct them through adaptive control, thereby maintaining error absorption capability while reducing the need for high initial manufacturing precision.
Data Source
AI summary
A control apparatus of a robot may include a state obtaining unit configured to obtain state observation data including flexible related observation data, which is observation data regarding a state of at least one of a flexible portion, a portion of the robot on a side where an object is gripped relative to the flexible portion, and the gripped object; and a controller configured to control the robot so as to output an action to be performed by the robot to perform predetermined work on the object, in response to receiving the state observation data, based on output obtained as a result of inputting the state observation data obtained by the state obtaining unit to a learning model, the learning model being learned in advance through machine learning and included in the controller.