Figures index

From

Angular Position Estimation of an Inverted Pendulum Using Low-Cost IMUs

Daisuke Aoyagi, Sukgi Choi

American Journal of Sensor Technology. 2018, 5(1), 1-6 doi:10.12691/ajst-5-1-1
  • Figure 1. Inverted pendulum and Data Acquisition Setup. We mounted BNO055 and MPU-9250 on an inverted pendulum. Two Arduino Uno clones served as I2C-to-USB/Serial translators. Data acquisition was triggered by data-ready signal of MPU-9250’s DMP running at 100 Hz. The distance from the pivot point of the inverted pendulum to the IMUs was: r=0.5 [m]
  • Figure 2. Sample Data Overview: Manually imposed motion included slow sweeping of about ±20°, sinusoidal-like swaying motion at 2‒4 Hz, and step-like movements. Note significant drift in the straight time integration (Cumtrapz) of the measured angular velocities
  • Figure 3. Sample Data Close-up View: Our implementation of Kalman Filter with the MPU-9250’s raw data produced the best result overall. The BNO055 exhibited noticeably larger lag and offset. The BNO055 Fusion exhibited noticeable non-smooth fluctuations, most notably in the middle row (roughly 4-Hz sinusoidal motion)
  • Figure 4. Raw sensor data corresponding to the onset of roughly 4-Hz sinusoidal-like motion (middle row of Figure 3, t=82.4‒83.4). The BNO055 exhibited a larger time lag than the MPU-9250 did. Note MPU-9250’s acceleration signal saturated at ±2g, which was the default full scale