IMU_dataset

A matlab dataset of various tests of inertial and magnetic sensors using an industrial robot as ground-truth, developed @ Instituto Politécnico de Leiria, Escola Superior de Tecnologia e Gestão.

Considerations of the dataset and tests

This repository intends to make available a dataset collected during an evaluation phase of a few fusion algorithms.

Allowing to compare the performance of fusion algorithms in a harsh dynamic environment with no prior knowledge of the magnetic field disturbances or linear accelerations caused by motion against some reference sensors available on the market.

The tests were carried out using an industrial robot that could retrieve the IMU unit orientation and position at 15 Hz, establishing the ground-truth.

All sensors signals were sampled at 100 Hz and interfaced to a PC via a custom wireless protocol between two MRF24J40 Microchip modules.

Both signals were synchronized in matlab time.

The tests were performed over 4 different paths at 6 different velocities with 8 different sensors. Run TestScript.m for more details.

6 different velocities:

• 150 mm/s (safe mode)
• 300 mm/s
• 500 mm/s
• 800 mm/s
• 1000 mm/s
• 1500 mm/s

4 different paths:

• path1 (yaw rotation - pitch rotation - roll rotation)
• path2 (linear movements)
• path3 (complex rotations)
• path4 (complex rotations)

4 sensors retrieving raw data:

• LSM9DS0 (gyro+acc+mag) 100Hz
• MPU9150 (gyro+acc+mag) 100Hz
• MPU6500 + RM3100 (gyro+acc+mag) 100Hz
• MPU6050 + RM3100(gyro+acc+mag) 100Hz

4 reference sensors retrieving orientation:

• PNI SENTRAL M&M Blue (gyro+acc+mag) 100Hz
• XSENS MTi-30 (gyro+acc+mag) 100Hz
• MPU6050 (gyro+acc) 100Hz
• MPU6050 (gyro+acc) 100Hz

Notes

Reference sensors/algorithms used:

• PNI SENTRAL M&M Blue Module runs a stateof-the-art patented KF algorithm performing continuous hard and soft-iron disturbance calibration and magnetic anomalies compensation;
• XSENS MTi-30 series runs a proprietary XSens Kalman Filter;
• MPU6500DMP and MPU6050DMP run an Invensense MotionFusion Algorithm (IMFA) with 6-axes (no usage of magnetometer data) integration.

Tested sensors:

• 9-axes integrated MPU9150 from Invensense;
• 9-axes integrated LSM9DS0 from STMicroeletronics;
• 9-axes integration with a 6-axes MPU6050 (Gyroscope+Accelerometer) and the 3-axes magnetic-inductive RM3100 (EVb) from PNICorp;
• 9-axes integration with a 6-axes MPU6500 (Gyroscope+Accelerometer) and the 3-axes RM3100 EVb.

Citations

If you use this software please cite one of the following papers:

1. Rasteiro, Miguel; Costelha, Hugo; Bento, Luis; Assuncao, Pedro, ”Accuracy versus complexity of MARG-based filters for remote control pointing devices," in Consumer Electronics - Taiwan (ICCE-TW), 2015 IEEE International Conference on, pp.51-52, 6-8 June 2015

2. Rasteiro, Miguel; Costelha, Hugo; Bento, Luis; Assuncao, Pedro, ”Low-complexity MARG Algorithms for Increased Accuracy in Space Pointing Devices”,in IEEE CE Workshop, March 11, 2015 Novi Sad, Serbia

3. Santos, Ricardo; Rasteiro, Miguel; Costelha, Hugo; Bento, Luis; “Assuncao, Pedro, Motion-based Remote Control Device for EnhancedInteraction with 3D Multimedia Content”, in Conference on Telecommunications (Conftele 2015), 17-18 September, 2015, Aveiro, Portugal

Acknowledgements

This work was developed within project HERMES, co-financed by European Union, COMPETE, QREN and Fundo Europeu de Desenvolvimento Regional (FEDER).