Robot Motion Prediction Control Using Decoupled State Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion prediction control systems for robots face challenges in generating control command values at a predetermined control cycle due to time requirements for updating internal states, especially when dealing with complex motions or multiple robots, leading to difficulties in synchronizing sensor information and maintaining accurate tracking of moving workpieces or robots.
Innovation Solution
A motion prediction control device and method that acquires sensor information independently of the robot's control cycle, predicts and updates the internal state of the object or robot using a Kalman filter-based state estimation algorithm, and generates control command values in real-time based on the latest internal state, allowing for real-time control without being influenced by state estimation processing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor information is acquired and internal state is updated within the robot control cycle, then control command values can be generated in real-time, but the processing time required for state estimation causes delays and reduces measurement precision
Solution Approach 1:
The patent applies preliminary action by performing state estimation processing independently of the control cycle. Sensor information is acquired and the internal state is updated in advance, before the control command generation is needed. This allows the control system to use pre-processed state information without waiting for real-time estimation during the control cycle, thereby eliminating processing delays while maintaining measurement precision.
2Measurement precision
If multiple sensors and measurement methods are used to determine position and posture, then measurement accuracy is improved, but synchronizing all sensor information becomes difficult and increases system complexity
Solution Approach 1:
The patent introduces an intermediary approach by using a unified state estimation framework that processes sensor information from multiple sources. Instead of directly synchronizing and comparing raw sensor data from different coordinate systems and measurement methods, the system uses an intermediate state representation that integrates all sensor information through a common estimation model, thereby reducing synchronization complexity while maintaining measurement precision.
3Reliability
If state estimation processing is performed at every control cycle, then control accuracy is maintained, but the calculation load increases significantly for complex motions or multiple robots
Solution Approach 1:
The patent applies preliminary action by completing state estimation processing before the control cycle begins. The internal state is updated in advance using sensor information, so that when the control cycle starts, the estimation is already complete. This separates the computationally intensive estimation process from the time-critical control execution, maintaining control accuracy while improving processing efficiency for complex scenarios.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
(A) One or both of an object and a robot (2) are measured by measuring units (12a and 12b) to acquire sensor information. (B) The internal state of one or both of the object and the robot (2) is predicted and updated by a state estimation unit (14) on the basis of the sensor information. (C) The internal state is stored by a data storage unit (16). (D) The robot (2) is controlled by a robot control unit (20). At (B), the internal state is updated by the state estimation unit (14) at an arbitrary timing that is independent of the control cycle of the robot. At (D), a prediction value necessary for controlling the robot is calculated by the robot control unit at a control cycle on the basis of the latest internal state stored in the data storage unit (16).