1. Decentralized Cooperative Trajectory Estimation for Autonomous Underwater Vehicles
Liam Paull1,2, Mae Seto2,3 and John Leonard1
1MIT CSAIL, 2University of New Brunswick, 3Defense R&D Canada

2. Challenges and Potential Benefits
High latency
Low bandwidth
Unacknowledged (broadcast)
Unreliable

3. Challenges and Potential Benefits
High latency
Low bandwidth
Unacknowledged (broadcast)
Unreliable
Potential Benefits:
Vehicles surface for GPS fix less frequently
Collected data more accurately localized through trajectory smoothing

4. Underwater Cooperative Localization
Time

5. Underwater Cooperative Localization
Time

6. Underwater Cooperative Localization
Time

7. Underwater Cooperative Localization
Time

8. Underwater Cooperative Localization
Time

9. Underwater Cooperative Localization
Time

10. Underwater Cooperative Localization
Time

11. Centralized Multi-AUV Pose Graph… …

12. Centralized Multi-AUV Pose Graph
GPS measurements … …

13. Centralized Multi-AUV Pose Graph
Compass measurements … …

14. Centralized Multi-AUV Pose Graph
DVL derived odometry … …

15. Centralized Multi-AUV Pose Graph…
Inter-vehicle range measurements …

16. Centralized Multi-AUV Pose Graph… …
Problem: Too much data to send through Acomms

17. Decentralized Multi-AUV Pose Graph…
New factor connects other vehicle nodes at times of contact

18. Advantages of Proposed Approach
Guaranteed connectedness of pose graph
Data throughput scales linearly with team size
Data throughput constant with time
No requirements on team hierarchy

19. 2 AUVS, One Surfacing for GPS

20. Different Packet Loss Rates