Transcript Slide 1

Mission Planning
• Multiple vehicle missions require the vehicles to
be in formation
• An initial formation has to be established before
the mission starts
• Vehicles cannot be deployed in formation
 Need to get the vehicles into initial formation
Path Generation’s General Problem
• Problem description
– Desired waypoints or manoeuvres
– Desired AUV formation
– Optimal energy expenditure or manoeuvring time
Path Generation’s General Problem
• Difficulties
– Go-to-formation manoeuvre (collision-free paths)
– Avoid (concave, static) obstacles
– Opt. 1 – Generate non-intersecting paths
– Opt. 2 – Allow for intersecting, “time-coordinated”
trajectories
– Meet energy and trajectory
constraints
Path Generation
• Avoid absolute timing (to cope with disturbances)
• Reaching the target positions simultaneously will
ensure the required initial formation.
Decouple spatial and temporal constraints!
Step 1. Produce paths p(t) without explicit time
constraints (polynomial function of t)
Step 2. Establish a timing law for t= t (t)
(polynomial)τ)
Inspired by the work of Isaac Kaminer and Reza Ghabcheloo
Path Generation
Decouple spatial and temporal constraints!
• To translate between t and τ use
(adopt a polynomial approach to h(t))
Step 3. Use an optimization of your choice for
criterion minimization.
Inspired by the work of Isaac Kaminer and Reza Ghabcheloo
Path Generation
• Polynomial equation
• Taken for all DOFs (extremely simple system!)
Inspired by the work of Oleg Yakimenko, NPS, USA
Multiple Vehicle Manoeuvres
• Most vehicles are underactuated and lack
hovering capabilities
 No station keeping at deployment time, no
station keeping after vehicles reach mission
starting positions
 Mission has to start on-the-fly
• Vehicles should automatically be driven from
deployment to mission start
Go To Formation
• 1st part of the overall manoeuvre
• Needs to take into account initial
deployment positions and
orientations
• Needs to ensure
simultaneous
arrival at
designated
positions and
orientations
Go To Formation
• Needs to avoid collisions until mission
control takes over
• Needs to consider
time, energy and
vehicle
constraints
• Method has been
extended to deal
with intersecting
paths.
Go To Formation
• How do we deal with “deviations from the
plan? (wind, currents, etc.)
• Use cooperative
path following!
(methods are
available - IST,
NTNU)
The Grex 2008 Azores Mission
• July 2008 in Horta, Faial (Açores)
• 5 Nations, 9 Institutions
• Successful sea trials of
Coordinated Path Following
and Coordinated Target
Tracking
The Grex 2008 Azores Mission
• The GREX System provides
– Central planning for teams
– Hardware/OS Abstraction
– Information and command flow accross multiple
vehicles
– Central and distributed mission planning
– Complex coordination capability
– Safety checks at several levels, transmitted via
status/emergency telegrams
The Grex Vehicle Architecture
The Grex 2008 Azores Mission
• Desired Path is
predefined
• DELFIMX follows a
mission consisting of
predefined #ARC and
#POINT (line) SVPs
Slide from Arvind Pereira, USC
Simple Path Following
The Grex 2008 Azores Mission
• Path already
provided
• Vehicles
coordinate with
leader vehicle
to do
coordinated
path-following
Slide from Arvind Pereira, USC
Coordinated Path Following
The Grex 2008 Azores Mission
• MVP implementation
to take vehicles to a
starting formation
• Will use
path/trajectory
planning algorithms
that do spatiotemporal deconfliction
• Obstacle avoidance is
also required
Slide from Arvind Pereira, USC
GoToFormation
The Grex 2008 Azores Mission
• Águas Vivas
moves around
while sending
DELFIMX its GPS
locations
• DELFIMX
estimates path
followed in GREX
SVMP format and
follows this path
Slide from Arvind Pereira, USC
Simple Target Tracking
The Grex 2008 Azores Mission
DELFIMX GREXmodule’s
estimator learns
path of Águas
Vivas and does a
coordinated pathfollowing with
Seabee AUV
using GREX
MVPs
Slide from Arvind Pereira, USC
Coordinated Target Tracking
The Grex 2008 Azores Mission
• Because GREX deals with multiple vehicles, a
Go-To-Formation behaviour is necessary for
each mission
• My part in the mission:
– Developed an algorithm for Go-To-Formation
– Helped developing the software for telegram
processing (communication on the vehicle side)
– Developed software to be ready to go to sea in
Sesimbra, September 2008
Grex 2008 Azores Mission: Results
• Path Following
Grex 2008 Azores Mission: Results
• Target Tracking
Future Work
1. Incorporate deconfliction in time
2. Go on to three-dimensional paths
3. Finally, think about on-line path planning
during mission runtime