Optimistic distributed simulation for a UAV flight control system

• US 6,498,968 B1
• Filed: 11/27/2001
• Issued: 12/24/2002
• Est. Priority Date: 11/27/2001
• Status: Expired due to Fees

• Current Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Original Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Inventors: Bush, Stephen Francis
• Primary Examiner(s): ARTHUR JEANGLAUDE, GERTRUDE
• Application Number:
  US09/994,448
• Time in Patent Office: 392 Days
• Field of Search: 701/1, 701/3, 701/33, 701/36, 701/66, 455/522
• US Class Current: 701/3
• CPC Class Codes: G05B 13/048 using a predictor

Abstract

A real-time distributed model-based predictive control method having a time maintenance mechanism that solves all state synchronization problems. Method steps include generating a message that predicts a system value corresponding to a future time, and executing a control process using the system value contained within said message, resulting in a process state. The process state is then stored in a state queue. The local process time is rolled back to the latest valid local process time prior to said message if the message precedes the current process state time. Rolling back comprises retrieving the system state values at that latest valid local process time from said state queue and re-executing the control process, using the system value contained within the message, and storing the new state in the state queue. The method also includes rolling back if, upon receiving a message containing an observed system value, said observed system value differs from its predicted value by more than a predetermined tolerance. The method is carried out on a distributed processor control system.

Claims

1. A method of real-time predictive control, said method comprising the steps of:

(a) generating a message that predicts a system value, said message having a receive time corresponding to a future time;

(b) executing a control process using the system value contained within said message, resulting in a process state having a local process time Corresponding to the receive time of said message;

(c) storing the process state in a state queue;

(d) rolling back the process state to the latest valid local process time prior to the receive time of said message, only if said receive time precedes the current process state time, said rolling back comprising;

(i) retrieving the system state value at said latest valid local process time from said state queue; and

(ii) performing steps (b) and (c), above, using the system value contained within said message.

2. The method of real-time predictive control according to claim 1, further comprising:

(e) sending messages containing the results of the control process, said messages having a receive time corresponding to the local process state time; and

(f) storing said sent messages in a send queue.

3. The method of real-time predictive control according to claim 2, wherein said rolling back step further comprises:

(iii) sending anti-messages which nullify any messages having a send time later than the receive time of the message which initiated the rollback.

4. The method of real-time predictive control according to claim 3, further comprising executing steps (d)(i) and (d)(iii) upon receiving an anti message having a receive time that precedes the local process time.

5. The method of real-time predictive control according to claim 2, further comprising:

(g) filtering said sent messages through a Kalman filter.

6. The method of real-time predictive control according to claim 1, further comprising:

(h) rolling back according to step (d), only if, upon receiving a message containing an observed system value, said observed system value differs from its predicted value by more than a predetermined tolerance.

7. A real-time predictive control system comprising a machine readable storage medium containing a program of instructions for causing a computer to perform the following steps:

(a) generating a message that predicts a system value, said message having a receive time corresponding to a future time;

(b) executing a control process using the system value contained within said message, resulting in a process state having a local process time corresponding to the receive time of said message;

(c) storing the process state in a state queue;

(d) rolling back the process state to the latest valid local process time prior to the receive time of said message, only if said receive time precedes the current process state time, said rolling back comprising;

(i) retrieving the system state values at said latest valid local process time from said state queue; and

(ii) performing steps (b) and (c), above, using the system value contained within said message.

8. The real-time predictive control system according to claim 7, further comprising a Kalman filter.

9. The real-time predictive control system according to claim 7 wherein the system being controlled is an unmanned air vehicle.

10. The real-time predictive control system according to claim 7 wherein the system being controlled is a plurality of unmanned air vehicles.

11. The real-time predictive control system according to claim 10 wherein said plurality of unmanned air vehicles comprises a swarm.

12. The real-time predictive control system according to claim 7 wherein said computer includes multiple processors operating in parallel.

13. The real-time predictive control system according to claim 12 wherein said multiple processors are distributed among a plurality of unmanned air vehicles which comprise a swarm.

14. A method of optimistic parallel real-time predictive control, comprising the steps of:

(a) providing a plurality of processors capable of working in parallel;

(b) distributing control processes among said plurality of processors;

(c) providing driving processes for generating messages that predict system values, said messages having a receive time corresponding to a future time;

(d) executing a control process with one of said plurality of processors, using the system values contained within said messages, in the order said messages arrive, resulting in a process state having a local process time corresponding to the receive time of said message;

(e) sending messages containing the results of said control process to other processors for further computation, said sent messages having a receive time corresponding to the local process state time;

(f) rolling back the local process time to the latest valid time before the receive time on an incoming message if said receive time precedes the current process time; and

(g) rolling back the local process time to the current wallclock time if an observed system value differs from its predicted value at the wallclock time by more than a predetermined tolerance.

15. The method of optimistic parallel real-time predictive control according to claim 14 further comprising controlling at least one unmanned air vehicle.

16. The method of optimistic parallel real-time predictive control according to claim 14 further comprising distributing said multiple processors among a plurality of unmanned air vehicles which comprise a swarm.