Robot Motion Control Using Learned Operating Force Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional robot operation control methods require significant human effort for programming and are time-consuming, necessitating a more efficient and flexible approach using machine learning to reduce creation time and adapt to various situations.
Innovation Solution
A robot system incorporating a robot, motion sensor, surrounding environment sensor, operation apparatus, learning control section, and relay apparatus, which uses machine learning to estimate and output calculation operating forces based on operator-operating forces, environment data, and operation commands, allowing seamless conversion of these forces into operation commands for the robot.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a program for operating a robot is created by human understanding and programming, then the robot can perform operations according to predefined instructions, but the creation and adjustment time of the program becomes excessively long
Solution Approach 1:
The patent replaces manual programming mechanisms with machine learning mechanisms. Instead of humans writing and adjusting programs, the system automatically learns operational patterns from demonstration data and generates control programs, thereby eliminating the time-consuming manual programming process while maintaining reliable robot execution
Solution Approach 2:
The robot system performs self-programming through machine learning. By capturing demonstration operations and automatically generating control programs from this data, the system serves itself in creating operational instructions, removing the need for external human programmers and significantly reducing program creation time
2Adaptability or versatility
If conventional programming methods are used to control robot operations, then the robot follows predefined instructions, but the system lacks flexibility to adapt to various situations
Solution Approach 1:
The patent transforms the static, predefined program into a dynamic learning system. The control program is no longer fixed but continuously adapts by learning from new demonstration data, allowing the robot to flexibly respond to various situations while the underlying machine learning framework provides the necessary computational structure
Solution Approach 2:
The machine learning-based control system serves multiple functions: it learns from demonstrations, generates control programs, adapts to new tasks, and handles various operational scenarios. This multi-functional approach replaces multiple specialized programming efforts with a single versatile learning system
3Productivity
If machine learning is used to construct a model for robot operation control, then program creation time is reduced and flexibility is improved, but the complexity of the control system increases
Solution Approach 1:
The patent introduces a machine learning model as an intermediary between demonstration data and robot control. This intermediary automatically processes raw demonstration data into structured control programs, bridging the gap between simple data collection and complex robot operations while managing system complexity through specialized learning algorithms
4Extent of automation
If a robot system with operation apparatus and learning control section is implemented, then autonomous and cooperative operation is enabled, but the complexity of the system architecture increases
Solution Approach 1:
The patent segments the control system into distinct functional modules: operation apparatus for input, learning control section for processing, model construction unit for learning, and program generation unit for output. This segmentation allows each component to specialize in a specific task, enabling autonomous operation while managing overall system complexity through modular design
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
A robot system (1) includes the robot (10), a motion sensor (11), a surrounding environment sensor (12, 13), an operation apparatus (21), a learning control section (41), and a relay apparatus (30). The robot (10) performs work based on an operation command. The operation apparatus (21) detects and outputs an operator-operating force applied by the operator. The learning control section (41) outputs a calculation operating force. The relay apparatus (30) outputs the operation command based on the operator-operating force and the calculation operating force. The learning control section (41) estimates and outputs the calculation operating force by using a model constructed by performing the machine learning of the operator-operating force, the surrounding environment data, the operation data, and the operation command based on the operation data and the surrounding environment data outputted by the sensors (11 to 13), and the operation command outputted by the relay apparatus (30).