Robot Teleoperation Using Bio-Signal Motion Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing tele-operation systems face delays in robot motion due to communication and control delays, leading to deteriorated operation efficiency and impaired operational feeling for the operator, especially when the operator's motion speed changes.
Innovation Solution
An operation system that uses a machine learning model to predict operator motion based on bio-signals, such as electromyograms or electroencephalograms, to control robot motion, reducing delays by anticipating the operator's intended actions and incorporating environmental context through image processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If tele-operation system transmits operator motion to remote robot through communication, then robot can be controlled remotely, but motion delay occurs due to communication and control delays
Solution Approach 1:
The system performs preliminary action by detecting bio-signals (EMG, EEG) that precede actual operator motion and using machine learning to predict the operator's intended motion before it occurs. This allows the robot to begin executing commands before the operator physically moves, compensating for communication and control delays inherent in remote tele-operation systems.
2Loss of time
If operator motion is predicted using traditional extrapolation techniques, then constant-speed motion can be compensated, but prediction accuracy deteriorates when motion speed changes
Solution Approach 1:
The system replaces traditional mechanical/extrapolation-based prediction methods with a machine learning model that processes bio-signal data. This substitution enables accurate prediction of operator motion even when speed changes occur, as the ML model learns patterns from bio-signals that precede motion, rather than relying on assumptions of constant velocity.
Solution Approach 2:
The system changes the parameters used for prediction from simple motion extrapolation to multi-parameter bio-signal analysis including electromyogram (EMG) and electroencephalogram (EEG) data. By incorporating these physiological parameters that directly reflect neural and muscular activity, the system achieves higher prediction accuracy across varying motion speeds.
3Measurement precision
If machine learning model predicts operator motion from bio-signal, then prediction accuracy improves, but system complexity increases
Solution Approach 1:
The system introduces an intermediary machine learning model that bridges the gap between raw bio-signal detection and robot control. This intermediary component processes the complex bio-signal data and translates it into predicted motion commands, managing the complexity internally while presenting a simplified interface for remote operation.
4Stability of the object's composition
If robot motion follows operator motion with delay, then system stability is maintained, but operation efficiency deteriorates
Solution Approach 1:
The system maintains stability while improving efficiency by performing preliminary action through bio-signal-based prediction. The robot executes commands based on predicted operator intent before the operator physically moves, reducing the effective delay without causing instability, as the prediction is based on physiological signals that reliably indicate intended motion.
Data Source
AI summary
An operation system and an operation method that may improve operability are provided. The operation system includes a circuitry configured to determine a predicted value of an operator motion after a predetermined prediction latency from a current time based on a bio-signal using a prescribed machine learning model, the bio-signal captured from the operator and a controller configured to control a motion of a robot based on the predicted value. The operation method includes a step of determining a predicted value of an operator motion after a predetermined latency from the current time based on a bio-signal using a prescribed machine learning model, the bio-signal captured from the operator, and a step of controlling a motion of a robot on the predicted value.