Master thesis: Experimental Modelling of Wrist Impedance Using Sparse Identification of Nonlinear Dynamical Systems (SINDy)

Project Description:

Precise measurement of wrist impedance is essential for creating safe and effective human-robot interaction systems. Ensuring safety in physical human-robot interactions is critical due to the direct contact between humans and robots. Robots can experience instability when encountering rigid external environments or when there is a mismatch between the intended human motion and the robot's movements. Therefore, understanding the dynamic properties of the human wrist is vital for improving stability and control in these interactions. This project aims to measure the dynamic impedance characteristics of the human wrist using a Universal Robot. The study will focus on both linear and rotational impedance, examining how different conditions, such as varying levels of muscle co-contraction and external perturbations, affect these characteristics. The research will also include the design of an ergonomic handle to ensure proper contact throughout the measurement process. Wearable Perturbation device also needed to be designed for different external perturbation scenarios. The results will contribute to a better understanding of the biomechanical properties of the human wrist and wrist, which is essential for designing advanced robotic systems and prosthetics.

 Tasks and Duties:

o   Literature Review:

·       Review existing studies on wrist and wrist impedance measurement.

·       Identify gaps in the current research, particularly in the context of rotational impedance.

o   Handle Design:

·       Design an ergonomic handle (e.g., ball shape, T-shape, donut, cylinder, or even a strap connection) that ensures proper contact throughout the measurement process.

·       Implement a proper sensor array for measuring force across the handle.

·       Test and iterate on the handle design to accommodate various hand sizes and grip strengths.

o   Experimental Setup:

·       Design a set of 3D trajectories considering safety features.

·       Use the joint angle controller of the UR10 for smooth dynamic movement.

·       Develop protocols (stepwise, periodic, etc.) for applying controlled perturbations to the wrist and wrist.

·       Design and fabricate a wearable perturbator for different parts of the elbow, forearm, and shoulder.

·       Incorporate additional accelerometers/positioning sensors for tracking the mentioned points.

o   Data Collection:

·       Conduct experiments to collect data on wrist and wrist impedance under various conditions, assessing the effects of frequency, velocity, trajectory, and mass.

·       Ensure consistent and accurate data collection by calibrating sensors and validating the setup.

o   Modelling procedure:

·       Extract impedance parameters (e.g., stiffness, damping, inertia) using system identification techniques (SINDy) to develop dynamic models of wrist impedance.

·       Tackle the potential challenges of SINDy modeling by tuning algorithm parameters. 

o   Model Validation:

·       Validate the developed models using various experimental tests and inputs (e.g., different types of perturbations).

·       Perform cross-validation by splitting the experimental data into training and validation sets. Use the training set to identify the model parameters and the validation set to test the model’s accuracy.

·       Conduct error analysis using residual analysis, root mean square error (RMSE), mean absolute error (MAE) and R-squared to quantify the model’s accuracy.

o   Documentation and Reporting:

·       Document the design process and experimental findings, Prepare presentation and scientific report.

Requirements:

o   Knowledge of robotics and experience with robotic manipulators (e.g., Universal Robot).

o   Familiarity with data acquisition and sensor integration.

o   Experience with programming languages such as Python or MATLAB for data analysis and modeling.

o   Strong background in biomechanics and mechanical modeling.

o   Interest in machine learning approaches (e.g., SINDy).

o   Ability to apply system identification and signal processing techniques.

o   Familiarity with 3D modeling (e.g., SolidWorks).

o   Strong analytical thinking skills.

Contact Person: Dr. Mohammad Sadeghi (mohammad.sadeghi@tuhh.de)

Desired starting date: As Soon As Possible