AUTONOMOUS CONTROL SYSTEM FOR SATELLITE FORMATION FLYING - PowerPoint PPT Presentation

1 / 33
About This Presentation
Title:

AUTONOMOUS CONTROL SYSTEM FOR SATELLITE FORMATION FLYING

Description:

Failure of a satellite = deterioration of the system, ... Quaternion feedback control. Satellite actuators. dynamics. Star, horizon. gyros and. Sun sensors ... – PowerPoint PPT presentation

Number of Views:208
Avg rating:3.0/5.0
Slides: 34
Provided by: kthana
Category:

less

Transcript and Presenter's Notes

Title: AUTONOMOUS CONTROL SYSTEM FOR SATELLITE FORMATION FLYING


1
AUTONOMOUS CONTROL SYSTEM FOR SATELLITE FORMATION
FLYING
  • K Thanapalan and S M Veres
  • School of Engineering Sciences,
  • The University of Southampton

2
Outline of the presentation
  • Formation Flying Concept
  • Activity at Southampton
  • The analytic (control) problems
  • The multiple agent architecture
  • Commands and messages
  • Conclusions on verification
  • Illustration

3
Formation Flying Concept
  • Idea Replace a large and unpractical satellite
    by several small and less expensive ones.
  • Advantages
  • Emulated system more flexible.
  • Failure of a satellite gt deterioration of the
    system, system performance, but does not
    jeopardise the whole mission.

4
4
  • Formation Flying Concept

concept pursued NASA and ESA and some other
agencies.
TS3links separate telescopes together to produce
one large image.
(NASA pictures)
5
  • Formation Flying Projects at the School of
    Engineering Sciences
  • University of Southampton
  • Constrained controller design project (optimal
    design software, EPSRC 178k)
  • Formation Flying electronics testing facility (5
    DOF frames into which control systems can be
    plugged into for testing, EPSRC 80k
  • Partners EADS Astrium and Aero-Astro

6
Laboratory Hardware 3 mockup satellites of 5DOF
7
  • Formation Flying challenges

Virtual forces acting on a follower craft
around a leader on natural orbit at 0,0,0
8
  • Formation Flying control challenges
  • Thruster resolution quantized forces and
    limited forces on electrical thrusters (lt10mN)
  • Thrusters generate torques not only
    translational forces. Possibly varying mass
    centre requires adaptation of controls to keep
    prescribed attitude while translation is
    performed by thrusters.
  • Constraints on thruster forces make some
    formations impossible

9
  • Formation Flying challenges
  • Collision avoidance while switching to new
    formation
  • Sensor deterioration or failures (gyros,
    horizon, sun, star sensors and stereo-vision
    system)
  • Actuator deterioration or failures (electric
    propulsion, reaction wheel, control moment gyro
    problems)
  • Autonomous control (there is no time for ground
    control to calculate necessary action before a
    satellite can fatally drift off from the cluster

10
  • Formation Flying do we need agent technology ?
  • Probably it could be done in object oriented
    programming, but ...
  • Each satellite needs to communicate with the
    others,
  • Each satellite has to make decisions on the
    basis of sensed and received information and
    commands.
  • Each satellite must be self-reliant in case of
    joint operation emergency caused by faults in the
    system. For instance must keep itself close to
    the cluster even if it cannot communicate with
    their peers.
  • Conclusion object oriented programming appears
    to be more tedious to deploy then agent oriented
    programming

11
What agent functionality is needed?
  • Cluster level agents to serve the cluster as a
    whole (not just an individual satellite) a
    mission manager, a cluster path planner and data
    fusion agents to combine all measurements to
    improve accuracy
  • Multi-agents on board each craft a control
    devices handler, a sensor handler and a local
    manager
  • Commands and data communication between mission
    manager and local managers
  • Commands and data communication between mission
    manager and local managers

12
The analytic control problems
  • Adaptive tracking of position and attitude
    various globally stable Lyapunov design based,
    adaptive control methods have been developed
    during the last few years
  • Fuel efficient path planning under no
    constraints (solutions derived by the Pontryagin
    principle.)
  • Input constrained solutions to fuel efficient
    formation control used for path planning.
  • Extended Kalman filtering for stereo vision
    based attitude and position estimation of each
    craft based on measurements taken by all
    instruments of the cluster of satellites in the
    formation.

13
Multi agent architecture for the cluster
For the whole cluster
Mission manager ...
Mission control (human)
Cluster path planner
Cluster data fusion
Maneuver Executor adaptive feedback controller
agent
Local manager
Controls handler
Local path planner agent
Sensors handler
Repeated on each satellite
14
Commands received by the mission manager
Possible Commands to mission manager
- install and wake all agents data agent team
data
- send devices health report data
managers list
- send general
report data agent team data
-
send general log data depth of detail

- consider and report on a proposed formation
data formation object, formation info
- supervise a new formation
data formation object, team data
- predict power sources
to maintain current formation
- maintain current
formation
15
Messages received by the mission manager
- From cluster planner checked conditions of
feasibility of proposed new formation - From
manager Sx started transfer to new formation
- From manager Sx finished transfer
to new formation, maintaining formation
- From manager Sx sensor Sx fault actuator
Ax fault - From manager
Sx software fault
messages
Install an agent, wake agents, send formation
commands make decision on reconfiguration,
organize emergency send report to mission
control, ask managers for physical health, ask
cluster planner, etc.
action set
16
Multi agent architecture for the cluster
For the whole cluster
Mission manager ...
Mission control (human)
Cluster planner
Cluster data fusion
Maneuver Executor adaptive feedback controller
agent
Local manager
Controls handler
Local path planner agent
Sensors handler
Repeated on each satellite
17
Messages received by the local manager
  • - maintain formation
  • - check devices
  • - send health report
  • - check conditions for new formation
  • transfer to position in new formation

Commands from Mission manager
  • received new path plan
  • executing transfer
  • unusual response of craft
  • adaptation-log of operations

Messages from Controller
18
Action set of local manager
Send message started transfer to new formation
Send message finished transfer to new
formation Send message sensor Sx fault
actuator Ax fault Monitor position Switch
controller scheme Send hardware health
report Send events log for last x minutes Predict
position of other satellites Ask for modified
path Compute emergence path Change path to
track
19
Multi agent architecture for the cluster
For the whole cluster
Mission manager ...
Mission control (human)
Cluster planner
Cluster data fusion
Maneuver Executor adaptive feedback controller
agent
Local manager
Controls handler
Local path planner agent
Sensors handler
Repeated on each satellite
20
Action set of cluster planner
  • new formation request
  • abort current planning process
  • limited capabilities description
  • send message working on transfer plan
  • send transfer_plan_complete
  • send requested transfer is not feasible

Incoming messages from mission manager
Outgoing messages
- computing new plan - idle
Action set
21
Multi agent architecture for the cluster
For the whole cluster
Mission manager ...
Mission control (human)
Cluster planner
Cluster data fusion
Maneuver Executor adaptive feedback controller
agent
Local manager
Controls handler
Local path planner agent
Sensors handler
Repeated on each satellite
22
Multi agent architecture for the cluster
  • track path x
  • use controller type x
  • current positions are x
  • current positions are x

commands from local manager
Outgoing messages
  • Control going all right
  • unusual craft response

Action set
  • operate controller x to track p
  • idle

23
Multi agent architecture for the cluster
For the whole cluster
Mission manager ...
Mission control (human)
Cluster planner
Cluster data fusion
Maneuver Executor adaptive feedback controller
agent
Local manager
Controls handler
Local path planner agent
Sensors handler
Repeated on each satellite
24
Sensor Data Fusion (SDF) agent
  • It is essentially an enhanced estimator for the
    satellite position and attitude in the local
    coordinate system that also has the ability to
    report sensor problems
  • It uses measurements obtained from sensors, if
    some data does not arrive or is faulty it is
    still able to come up with an estimate
  • For each sensor there is a handler agent that is
    capable to monitor the reliability of the sensor
    (solid state gyros, stereo vision, star, Earth
    and Sun sensors) and reports malfunction to SDF

25
Sensor Data Fusion (SDF) agent
Incoming messages
receive all gyro data receive all sun sensor
data receive all star data receive all vision
data receive all horizon data Receive all GPS
data
Action set
Estimate all positions and attitudes Broadcast
all positions and attitudes Report detected
malfunction
26
Stereo vision system
27
Joint control of attitude and position
28
Further details of joint attitude and position
control of a single satellite

29
Cluster level control

30
A simple jet actuator configuration
31
Progress so far
  • We are still working on increasing messaging
    vocabulary
  • We have coded the adaptive feedback control
    agents
  • We are testing the SDF agent partly on real and
    partly on simulated data.
  • The position and attitude data communicator
    essentially complete
  • We have built a laboratory testing hardware
    including 3 satellite models.

32
  • Our conclusions and experience
  • In our EPSRC projects it was not a requirement
    that we have to use agent technology but we have
    found them a useful technique that speeds up
    developing a reliable control structure.
  • Verification We are using a minimal complexity
    agent architecture that does the job it can be
    modelled by a finite state machine if abstract
    the behaviour of the 4 analytic modules (1)
    path planning, (2) adaptive controller (3) data
    fusion filter
  • If we assume that the analytic modules behave
    as predicted by physical laws and the floating
    point arithmetic based algorithms are 100
    reliable then the system is verifiable for a
    given set of failure modes.
  • Agents are useful for functional modularity,
    reliability and maintainability of software.
    Encapsulation of behaviour allows easier testing
    and checking logical consistency of the whole
    system, i.e. verification.

33
Thank you
Write a Comment
User Comments (0)
About PowerShow.com