A PSO-ML-LSTM-based IMU state estimation approach for manipulator teleoperation.

Area of Science:

• Robotics
• Artificial Intelligence
• Sensor Fusion

Background:

• Teleoperation enables remote execution of hazardous tasks but is hindered by signal noise and cumulative errors from inertial measurement units (IMUs).
• Accurate state estimation of IMUs is crucial for maintaining precise control in robot teleoperation systems.
• Existing methods struggle to effectively compensate for the drift and cumulative errors inherent in IMU data.

Purpose of the Study:

• To develop and validate an advanced IMU state estimation method to mitigate cumulative errors in robot teleoperation.
• To enhance the performance and reliability of teleoperation systems by addressing IMU-induced inaccuracies.
• To improve the safety and efficiency of robots performing tasks in hazardous environments through precise remote control.

Main Methods:

• Established a motion mapping model for human arm and a 7-DOF robotic arm using global configuration and hybrid mapping.
• Constructed an IMU pose state estimation model utilizing particle swarm optimization (PSO) and modulated long short-term memory (ML-LSTM) neural networks.
• Trained the estimation model with initial data from multiple IMUs and handling handles.

Main Results:

• The proposed PSO-ML-LSTM algorithm demonstrated significant effectiveness in eliminating the impact of IMU cumulative errors.
• Comparative experiments confirmed the superior performance of the developed state estimation model over conventional approaches.
• The hybrid mapping and advanced neural network model accurately described the influence of IMU errors on teleoperation.

Conclusions:

• The PSO-ML-LSTM approach provides a robust solution for IMU state estimation in robot teleoperation.
• This method significantly improves the accuracy and reliability of teleoperation, enabling safer execution of tasks.
• The findings pave the way for more dependable and high-performance robotic systems in hazardous environments.