Guidance:

Dynamic parameter identification code for rokae xmate manipulator based on MATLAB, including excitation trajectory optimization, LSM method, and Recursive Newton-Euler Algorithm.

Project Stucture and Description:

Dynamics:

• Formulate robot dynamic model through Newton-Euler method.
• Linearize robot dynamic model to obtain regressor.
• Apply QR decomposition to obtain minimal regressor.
• Convert minimal parameter set Pmin to standard set.
• Verify error between observation and NE-based estimation.
• Deduce robot dynamic equation item.

See ./dynamics/README.md and run_dynamics.m for details.

Excitation:

• optimize excitation trajectory based on cond of minimal regressor.
• matrixized constraints for trajectory optimization: |q| < qmax, |qd| < qdmax, |qdd| < qddmax, q0=qn=q_init
• plot figures of {q, qd, qdd, cart_pos} and also animation.
• cpp scripts for running excitation trajectory based on rci client.

See ./excitation/README.md and run_optimize.m for details.

Filtering:

• Downsample observation data to assigned size.
• Apply butterworth and zero-phase filter to q, qd, qdd, tau data.
• Plot figures of raw and filtered data.

See ./filtering/README.md and run_filtering.m for details.

Identify:

• Apply LSE (least square estimation) to figure out minimal param set.
• Verify error between observation and estimation by min regressor.

See ./identify/README.md and run_identify.m for details.

Usage scenarios:

Identification pipeline:

1. Derive robot dynamics, regressor and minimum paramset: run_dynamics.m PART-IA and PART-IB.
2. Optimize excitation trajectory: run_optimze.m.
3. Data filtering and processing: run_filtering.m.
4. Estimate minimum paramset using LSE: run_identify.m.
5. Map minimum paramset to standard paramset: run_dynamics.m PART-II.
6. [OPTIONAL] Test the performance of identified dynamics model with standard paramset: run_dynamics.m PART-III.

Notes:

1. Keep System of Units consistent throughout the project (mm and Nmm).
2. Additional adjustment of virtual mass in dyn_mapping_Pmin2P.m is needed for better paramset mapping.

Excitation Trajectory Optimization:

Obtain min regressor matrix in \dynamics and then turn to \excitation.

Validation Error Verification:

Copy raw sensor data in \filtering and then turn to \dynamics.

References:

[1] Yanjun Liu. "Study on Parameter Identification for Rokae XB4," 2019. link.

[2] Craig, John J. "Introduction to Robotics," 2005.