DriftLess Technology to improve inertial sensors

Size: px

Start display at page:

Download "DriftLess Technology to improve inertial sensors"

Griffin Foster
6 years ago
Views:

1 Slide 1 of 19 DriftLess Technology to improve inertial sensors Marcel Ruizenaar, TNO marcel.ruizenaar@tno.nl

2 Slide 2 of 19 Topics Problem, Drift in INS due to bias DriftLess technology What is it How it works History INS integrated navigation Field tests/examples Accelerometer bias estimation Airborne test, gyro bias estimation Indoor AGV test, gyro bias estimation Conclusions

Slide 3 of 19 Drift problem (1/3) INS based systems use inertial sensors Accelerometers measure linear acceleration Gyroscopes measure angular velocity Inertial sensors have small

3 Slide 3 of 19 Drift problem (1/3) INS based systems use inertial sensors Accelerometers measure linear acceleration Gyroscopes measure angular velocity Inertial sensors have small unknown offsets (biases) Due to numerical integration, bias results in ever growing error (drift) Drift in attitude estimation. Drift in velocity estimation. Drift in position estimation.

4 Bias [mg] Bias [deg/hr] Slide 4 of 19 Drift problem (2/3): how large is the bias? Example bias of low-cost MEMS gyroscopes/accelerometers (MPU6050) 14 hour of measurement data At constant temperature Turn-on bias already removed In-run bias variations 1 Bias accelerometers 200 Bias gyroscopes; rotor set A1x A1y A1z A2x A2y A2z G1x G1y G1z G2x G2y G2z Time [hr] Time [hr]

5 Slide 5 of 19 Drift problem (3/3) How big is the problem, using typical low-cost MEMS sensors? 5 mg accelerometer bias 200 deg/hr gyro bias (~1 mrad/s) * Integration time 10 seconds 2.5 m 23 m 10 mrad/0.6 deg 1.7 m Integration time 30 seconds 30 mrad/1.7 deg 45 m P drift = 1 2 b at 2 P drift = 1 6 b gg 0 t 3 *1 mrad (~0.05 deg) attitude error 1 mg acceleration error. Correcting the bias of the gyros has major priority!!! Very limited use of MEMS based INS-only systems

6 Slide 6 of 19 DriftLess technology (1/3) Estimates biases of sensors that measure a vector quantity Inertial sensors (accelerometers & gyroscopes) Magnetometers Other sensors 1D, 2D or full 3D Almost independent of temperature Bias estimation and compensation continuously during use/operation. All fixed and varying biases removed, leaving only white noise (theoretically). Bias compensation of all sensors can be combined in a single set-up. Patent pending TNO-technology

7 Slide 7 of 19 DriftLess technology (2/3): basic principle DriftLess box

8 Slide 8 of 19 DriftLess TM, how does it work (3/3) R 1 1 s1 = R 1 1 R 1 1 x + R 1 1 b 1 + R 1 1 n 1 R 2 1 s2 = R 2 1 R 2 1 x + R 2 1 b 2 + R 2 1 n 2 y y = RR 1 b R b R 1 1 n 1 R 1 1 b 1 R 1 2 b 2 + n 2 n 2

9 Slide 9 of 19 Historic development of DriftLess TM 2010:Experimental setup (2D Static) 2010: Proof of concept TNO IenT (2D Static) 2011 Experimental setup (3D dynamic) 2012: NTP DriftLess TM 2013: DriftLess TM for Award 2014: DriftLess for AGV (in progress)

Slide 10 of 19 Integrated navigation systems Gyro biases only indirectly observable Attitude errors equivalent with accelerometer biases Accuracy bias estimation depends on IMU noise Bias stability

10 Slide 10 of 19 Integrated navigation systems Gyro biases only indirectly observable Attitude errors equivalent with accelerometer biases Accuracy bias estimation depends on IMU noise Bias stability Measurement noise MEMS gyros: long estimation time Combine with DriftLess Biases already estimated IMU b IMU 0 Drift- b IMU = 0 Less 9 state filter Less filter states (computationally more attractive) Allows for longer unaided periods 15 state filter

11 bias [mg] Temperature [ C] Slide 11 of 19 Example 1: accelerometer bias estimation Output of 1 accelerometer 1 hour, corrected with DriftLess technology 4 mg bias change due to temperature rise of 0.5 degree With DriftLess, residual bias < 0.1 mg 5 True vs estimated bias; sensor 2 25 Temperature mg deg -1 True bias Estimated bias time [min] time [min]

12 Slide 12 of 19 Example 2: Field test with DriftLess (1/3) INS/GNSS/DriftLess processing IMU s & GPS receiver inside box Processing afterwards in the lab Triangulate PR to get position Process PR + Doppler to get blended INS/GPS solution

clock bias/drift) No ground truth available Simulated GPS outages Calculated

13 Slide 13 of 19 Example 2: Field tests with DriftLess (2/3) Main goal to improve SAR imaging Tightly coupled INS/GNSS configuration 9+2 states (pos, vel, att + clock bias/drift) No ground truth available Simulated GPS outages Calculated equivalent gyro bias Stationary During flight P drift = 1 6 b gg 0 t 3 P drift = 1 2 b at 2

Equivalent gyro bias ~ 20 deg/hr INS/GNSS 15state: >100 deg/hr 8 6 4 2 0-2 Roll Pitch -4-6

14 Attitude [deg] [deg] Slide 14 of 19 Example 2: Field tests with DriftLess (3/3) Drift observed during: Airplane stationary (~8 min outage) Airplane in flight (~90 sec outage) Equivalent gyro bias ~ 20 deg/hr INS/GNSS 15state: >100 deg/hr Roll Pitch -4-6 Zoomed-in Time [min] Time [min] P drift = 1 6 b gg 0 t 3 P drift = 1 2 b at 2

15 [deg] Slide 15 of 19 Example 3: DINS indoor (1/3) INS attitude reconstruction 200 INS attitude Time [s]

16 [deg] Slide 16 of Example 3: INS-only indoor (2/3) Tracking attitude INS attitude + attitude based on accelerometers Accelerometer biases accurate to 0.1 mg with DriftLess if stationary Accelerometer attitude measurement accurate to 0.01 deg Use as baseline during stationary conditions Observed gyro bias ~ 1 deg/hr Accelerometer bias shift ~2mg Time [s]

17 [deg] Slide 17 of 19 Example 3: DINS indoor (3/3) INS attitude reconstruction INS attitude + attitude based on accelerometers deg Equivalent gyro drift: 1 deg/hr ARW: ~0.02 deg Time [s]

18 Slide 18 of 19 Conclusions DriftLess capable of correcting nearly all fixed and varying bias in dynamic situations. Residual gyro bias ~1 deg/hr demonstrated Residual accelerometer bias <0.1 mg Integrated systems more accurate when combined with DriftLess Further work Miniaturization of the set-up, current size: 10x10x10 cm Upcoming: first miniaturization step: 4x4x4 cm Application in UAV s, ROV s or other small automated platforms Future miniaturization: 1x1x1 cm Application in Smart Phone applications (military or civil) or navigation equipment (TomTom, Garmin) Far future: rotation mechanism implemented on chip Civil Smart phone applications

19 Slide 19 of 19 Drawbacks, limitations and unknowns Due to imperfect calibration, residual bias depends on input signal Currently no experience with mechanical limitations Mechanical: how about life time expectancy Mechanical: shock resistance?

Inertial Navigation Systems

Inertial Navigation Systems Kiril Alexiev University of Pavia March 2017 1 /89 Navigation Estimate the position and orientation. Inertial navigation one of possible instruments. Newton law is used: F =