Autonomous Maritime Vehicle Systems @ Virginia Tech

1
Autonomous Maritime Vehicle Systems _at_ Virginia
Tech

Wayne Neu, Craig Woolsey, Dan Stilwell, Chris
Wyatt, Mike Roan Contact Dan
Stilwell stilwell_at_vt.edu [phone]
2

• Autonomous Vehicles
• High-Speed AUV
• 475 AUV
• VT ASV
• Fundamental Research
• Dynamics and control
• Multi-vehicle cooperation
• Data fusion
• Stochastic mapping
• Applied Research
• Environmental adaptive sampling
• Control design
• Distributed navigation
• Distributed signal processing

• Contributors
• Wayne Neu (AOE)
• Craig Woolsey (AOE)
• Dan Stilwell (ECE)
• Chris Wyatt (ECE)
• Mike Roan (ME)

3
High-Speed AUV

• Engineering highlights
• No passive roll stability
• requires active roll control
• 50 heavier than displacement
• sinks fast when not moving
• Nose-down hover when not in flight
• Virginia Tech activities
• Propulsion
• Hydrodynamics
• Guidance/control
• Electronics/software
• Flight testing
• Development Costs 350K
• Development time 10 months

Parameter Value (units)
Length 39 (inches)
Diameter 3 (inches)
Displacement 7.84 (lbs)
4
HSAUV Launch
Neutrally ballasted vehicle at high speed
5
Heavy Ballast, AUVFest 2007
Animation of data from AUVfest June 7, 2007 675
ft. run at 10 Knots (40 sec)
6
Active Roll Control

• Two independent props provide thrust roll
  control
• Allows orientation control in hover

7
475 AUV

• Design goals
• Rapid algorithm development
• Low-cost (9K)
• Orthodox hardware/software
• Features
• Acoustic comms and nav
• Client/server software architecture
• Removable mast

PARAMETER SPECIFICATION
Length 34 inches
Diameter 4.75 inches
Mass 18.3 lbs.
Stability passively stable in pitch and roll
Propulsion brushless DC motor with encoder feedback
Computer control system x86 compatible
Software LINUX OS, MOOS-like database server architecture utilizing TCP/IP client/server connections
Power 200 watt-hour lithium polymer battery stack
Communications 900MHz RF modem with ¼ wave antenna Wi-Fi with external antenna WHOI micromodem for acoustic communication
Data storage 1GB compact flash
Moving mass 9.9 lbs electronics carriage with battery stack moves 0.5 inches longitudinally
8
475 AUV

• CTD/DO probe
• Towed array (on-going)
• Blueview FLS (on-going)
• Magnetometer (on-going)

Payloads
9
Towed array
8 Piezoceramic Cylindrical Broadband Hydrophones
2cm
All analog and digital electronics
Ethernet to AUV
10
Adaptive Environmental Sampling

• Adaptive transects
• Create plume map, or boundary map, or track a
  boundary
• Utilize a plume indicator function

temperature
plume indicator function
Boundary track
Temperature alone does not predict outflow
Plume indicator function more clearly shows
outflow
11
Small plume localization/mapping
12
Multi-Vehicle Coordination
Key Theoretical Challenges
Closed-loop data fusion and control
Communication

• Control and estimation are coupled
• Unwanted coupling matters for fast and/or
  bandwidth-limited systems

• Sparse and time-varying networks topologies
• Low bandwidth (80 bits/sec!?)
• Latencies

Stilwell, D. J., Bollt, E. M., Roberson, D. G.,
2006, "Sufficient Conditions for Fast Switching
Synchronization in Time-Varying Network
Topologies," SIAM J. Applied Dynamical Systems,
vol. 6, no. 1, pp. 140-156. Porfiri, M.
Stilwell, D. J., Bollt, E. M., Skufca, J. D.
2007, Random Talk Random Walk and
Synchronizability in a Moving Neighborhood
Network, in Physica D, in press. Porfiri, M.,
Roberson, D. G., Stilwell, D. J., Tracking and
Formation Control of Multiple Autonomous Agents
A Two-Level Consensus Approach, Automatica, in
press.
13
Sparse Stochastic Networks
Expected value of network

• Results
• Notion of network time constant
• Relationship between network time-constant and
  time-constant of underlying dynamics
• Proximity graphs, controlled Markov chains

Porfiri, M., Stilwell, D. J., "Consensus Seeking
over Random Weighted Directed Graphs," in
IEEE Transactions on Automatic Control, (in
press) Porfiri, M., Stilwell, D. J., Bollt, E.
M., Synchronization in random weighted directed
networks, IEEE Transactions of Circuits and
Systems I (in press), and ACC 2007. Porfiri,
M., Roberson, D. G., Stilwell, D. J., Fast
switching analysis of linear switched systems
using exponential splitting, SIAM Journal of
Control and Optimization (in review) and ACC
2006.
14
Sparse Stochastic Networks

• Data fusion with observer structure
• (e.g. Kalman filter)
• Block-diagonalization for certain network
  topologies
• Two-level consensus framework
• Traditional data fusion
• ?? (new effort)

Closed-loop data fusion and control
Porfiri, M., Roberson, D. G., Stilwell, D. J.,
2006, "Environmental Tracking and Formation
Control of a Platoon of Autonomous Vehicles
Subject to Limited Communication," Proceedings of
the IEEE Int'l. Conf. on Robotics and Automation,
Orlando, FL. Roberson, D. G., Stilwell, D. J.,
"Decentralized Control and Estimation for a
Platoon of Autonomous Vehicles with a Circulant
Communication Network," Automatica, (in review)
and ACC 2006.
15
Example Solutions/Applications
AUVFest 2007
Tracking (vector field)
Tracking (scalar field)
16
Autonomous Surface Vehicle

• Capabilities
• Long-endurance (4 days)
• Robust
• 250lb payload
• Goal
• Autonomous navigation/mapping in unstructured
  environments
• Sensors/Electronics
• Laptop(s) for control and image processing
• Wifi (mesh network)
• Gyro-stabilized pitch, roll, heading
• Omni-directional camera (stereo on going)
• WAAS-GPS
• Water flow velocity (DVL)
• Depth
• CTD/DO

17
Navigation/mapping in unstructured environments
Feature detection, classification, localization
stochastic map generation
Path planning

• Challenges
• Mapping and path planning should be independent
  of sensor
• Many false features in maritime environment
• Current focus
• Moving obstacle detection and tracking
• Efficient distributed mapping and path planning
  for multiple vehicles

18
Feature map generation
19
Final feature map
20
Numerical and experimental AUV modelling for
control and design
Field data
Numerical models
21
Nonlinear Control Of Advanced AUVs

• Energy-based nonlinear control of streamlined
  AUVs
• Exploit intrinsic agility of vectored thrust
  vehicles.
• Enhance operability in dynamic, unstructured
  environments.
• Optimal motion planning for underwater gliders
• Analytically characterize lateral-directional
  maneuvers.
• Leverage results from nonholonomic robot control.

1G. DSpain (MPL/SIO) P. Brodsky (APL/UW) will
speak about Liberdade/XRay development at 845
AM.
22
Control of Slender, Agile AUVs
Objective Large-envelope AUV control
Approach Potential energy shaping
Takegaki Arimoto, 1981. Leonard, 1996.
23
Some Results...
Potential shaping yields almost global asymptotic
stability.1
Animation generated using VRMLPlot (C. Sayers).
Vehicle prototype by J. Graver.
1Woolsey, IEEE Conf. Decision Control, Dec. 2006
24
Step 1 The Steady Turn (A Regular Perturbation
Problem)
Simulations use Slocum dynamic model given by
Bhatta, 2006.
25
Step 2 Optimal Motion Planning (Dubins Car)

• The minimum time path at constant speed and
  maximum L/D is the minimum potential energy path.
• Rich, current literature on path planning for
  Dubins car1
• Point-to-point problems with specified final
  heading
• Point-to-point problems without final heading
• Multiple waypoint (travelling salesman) problems

1See, for example, Savla, Bullo, Frazzoli,
2006 Ma Castanon, 2006. Also see Sussmann
Tang, 1991 Boissonnat et al, 1992.