AUV Self-Localization Using a Tetrahedral Array and Passive Acoustics

Transcription

1 AUV Self-Localization Using a Tetrahedral Array and Passive Acoustics Nicholas R. Rypkema Erin M. Fischell Henrik Schmidt

2 Background - Motivation Motivation: Accurate localization for miniature, low-cost AUVs - Testbed to experiment with risky behaviors - Lower the barrier toward multi-auv research - Distributed sensing to capture spatial/temporal variation of ocean processes - Virtual dynamic acoustic arrays 1

3 Background - Motivation Motivation: Accurate localization for miniature, low-cost AUVs - Testbed to experiment with risky behaviors - Lower the barrier toward multi-auv research - Distributed sensing to capture spatial/temporal variation of ocean processes - Virtual dynamic acoustic arrays Size, power and cost constraints prevent the use of typical INS sensors Inertial navigation quickly diverges using prop. speed and MEMS IMU 1

4 Background - Approach Constrain unbounded growth in localization error using acoustics Standard approach: long baseline (LBL) - Multiple acoustic transponders, two-way travel-time (TWTT) + trilateration - TWTT requires AUV acoustic transmission power hungry, does not scale - Unwieldy to setup 2

5 Background - Approach Constrain unbounded growth in localization error using acoustics Standard approach: long baseline (LBL) - Multiple acoustic transponders, two-way travel-time (TWTT) + trilateration - TWTT requires AUV acoustic transmission power hungry, does not scale - Unwieldy to setup Our approach: one-way travel-time inverted ultra-short baseline (iusbl) - Single acoustic transmitter, inverted USBL receiver calculates range + azimuth from AUV to transmitter - Acoustically passive on AUV localize multiple vehicles with single transmitter 2

6 System Overview Beacon: all COTS components / AUV + receiver: first prototype mini, very-low cost SandShark AUV from Bluefin Robotics 3

7 System Overview Array Characteristics Tetrahedral array - minimum number of elements to provide near-consistent main lobe beamwidth regardless of acoustic steering angle Increase in frequency narrows main lobe beamwidth but increases sidelobe level 4

8 System Overview Array Characteristics Array size constrained by end-plate size, operating frequency selected to balance trade-off between main lobe beamwidth and sidelobe magnitude khz up-chirp selected (within Lubell rated frequency response) Example inwater up-chirp 5

9 Matched Filtering Range Estimation OWTT estimated via matched filtering - i.e. convolution of signal received on each hydrophone with stored up-chirp replica Matched filter output is combined using sum of product of unique pairs and converted to range estimate signal by scaling sample numbers by c/f s range (m) 6

10 inclination (deg) Beamforming Azimuth/Inclination Estimation AUV-beacon azimuth/inclination estimated via phased-array beamforming - i.e. summation of phase-shifted array signals over grid of look-angles Output is 3D array [no. inclinations no. azimuths n] - convert to azimuthinclination heatmap by selecting max n slice 7 azimuth (deg)

11 Acoustics vs. GPS Statistics System implemented with boom-mounted array on WAM-V ASV to allow for range/azimuth comparison against dual-antenna GPS Argmax of range signal and azimuth-inclination heatmap vs GPS statistics indicate std. dev. of ~2.7 m in range and ~6.2 deg in azimuth (lower accuracy bound due to motor noise likely to be better) 8

12 Particle Filter Localization Matched filtering + beamforming + IMU (heading) provide an instantaneous estimate of AUV position Acoustic propagation exhibits properties difficult to counteract multipath, interference result in non-gaussian distributions, measurements are multi-modal Motivate the use of a particle filter/sequential Monte-Carlo fuse range signal and beamformed heatmap with MEMS IMU heading and prop speed 9

13 SandShark Experiments Field experiments in Charles River with Sandshark AUV - Pre-programmed 1200 s mission with AUV running back-and-forth along dock for 70 m - 2 m depth, 1.4 m/s speed, periodic GPS surfacing Real-time onboard self-localization, particle filter solution fed-back to AUV for closed loop control One run included two commercial REMUS LBL acoustic transponders for verification 10

14 SandShark Results Dead-reckoning (black) large jumps upon surfacing, particle filter (green) consistent with GPS, red circles indicate GPS surfacing locations 11

15 SandShark Results Comparison to commercial LBL (yellow): still subject to acoustic errors/outliers Particle filter has better agreement with LBL than dead-reckoning Difference in trajectory vs. LBL: - DR (mean): 5.04 m - PF (mean): 3.48 m 12

16 Bluefin-21 Preliminary Experiments Identical system has been implemented on our Bluefin-21 AUV Macrura Real-time relative beacon localization can be used for behaviors such as beacon homing useful for under-ice/arctic operations Preliminary experiments in Massachusetts bay ship-mounted beacon used for AUV return-to-ship behavior 13

17 Bluefin-21 Preliminary Results Inertial navigation (red) DVL-aided INS, ship GPS (green) anchored at ~(-300, -280), particle filter beacon estimate (blue circle) range offset error 14

18 Summary Implementation and demonstration of a single-beacon acoustic navigation system that is - Low cost, easily deployable, and can localize a mini, low-cost AUV without DVL - Passive on the AUV, enabling the localization of multiple vehicles - Performs in real-time on a cheap embedded system - Enables the possibility of new multi-auv behaviors - Beacon homing/following - Multi-AUV surveying and control via a beacon-enabled leader vehicle Future work: Multi-AUV experiments localization of multiple vehicles, behaviors for distributed sampling, coherent acoustic processing 15