Robot Control Device Optimizing Trajectory Following Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In trajectory following control of robot arms, accurately identifying a dynamic model is challenging, making it difficult to achieve high-accuracy performance, and existing methods like repetitive learning require significant time and effort, leading to mechanical wear.
Innovation Solution
A robot control device comprising a log acquisitor, a first adjuster, and a second adjuster that acquires operation data, optimizes physical parameters for feedback control, and calculates trajectories for feed-forward control using a dynamic model, reducing errors and improving performance through off-line simulations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If model-based control is used to achieve high-accuracy trajectory following, then following performance is improved, but it requires accurate dynamic model identification which is difficult to obtain
Solution Approach 1:
The patent performs preliminary identification of dynamic model parameters before actual trajectory following control. By pre-identifying inertia, friction, and gravity parameters through separate experiments, the system prepares accurate model data in advance, avoiding the need for complex real-time identification during operation.
Solution Approach 2:
The patent divides the control system into two independent parts: model-based feedforward control using pre-identified parameters and reactive feedback control. This segmentation allows each part to be optimized separately - the feedforward part uses accurate pre-identified parameters for high accuracy, while the feedback part handles disturbances without requiring complex real-time identification.
2Device complexity
If repetitive learning control is used to avoid dynamic model identification, then model identification complexity is reduced, but it requires repeated robot operations which consume time and cause mechanical wear
Solution Approach 1:
The patent performs preliminary identification of dynamic model parameters through dedicated experiments before normal operation. By completing parameter identification in advance using safe test operations, the system avoids the need for repeated learning operations during actual production, significantly reducing time loss and mechanical wear.
Solution Approach 2:
The patent creates a virtual model (copy) of the robot arm using identified parameters, including inertia matrix, friction coefficients, and gravity vectors. This copied model can be used for simulation and control calculations without requiring physical repeated operations, allowing virtual testing and optimization before actual deployment.
3Loss of time
If off-line simulator is used for supervised learning, then repetitive operations are reduced, but the simulator requires accurate dynamic model identification
Solution Approach 1:
The patent performs preliminary identification of all necessary dynamic model parameters (inertia, friction, gravity) before constructing the off-line simulator. By preparing accurate parameter data in advance through separate identification experiments, the simulator can be built with high accuracy without requiring repeated operations during the identification phase.
Solution Approach 2:
The patent creates an accurate virtual copy of the robot arm in the off-line simulator using pre-identified parameters. This copied model reproduces the real robot's dynamics accurately, allowing supervised learning and trajectory optimization to be performed virtually without affecting the physical robot, thus reducing time loss while maintaining accuracy.
4Reliability
If safe parameters are selected during repetitive learning, then robot arm safety is improved, but following performance and learning speed deteriorate
Solution Approach 1:
The patent performs preliminary identification of dynamic parameters using safe, limited-range test operations before deploying the model for actual trajectory following. By completing parameter identification in advance under safe conditions, the system can then use the identified model for high-accuracy control without requiring repeated safe test operations, thus maintaining safety while achieving high performance.
Solution Approach 2:
The patent introduces an off-line simulator as an intermediary between parameter identification and actual robot operation. The simulator allows accurate dynamic model identification and trajectory optimization to be performed virtually using safe parameter ranges, while the physical robot only needs to execute pre-planned trajectories, thus maintaining safety without compromising following accuracy.
Data Source
AI summary
A robot control device includes a log acquisitor, a first adjuster, and a second adjuster. The log acquisitor is configured to acquire operation data of a robot arm which has been operated by making a target portion of the robot arm follow a predefined target path under a feedback control. The first adjuster is configured to adjust, based on the operation data acquired by the log acquisitor, a first physical parameter for calculating a trajectory of the target portion, to reduce errors between the predefined target path and positions of the target portion. The second adjuster is configured to calculate, based on the first physical parameter adjusted by the first adjuster, the trajectory of the target portion, the second adjuster that is configured to adjust, based on the trajectory calculated by the second adjuster, a second physical parameter to be used for a feed-forward control for controlling the robot arm.