Software-Enabled Control

1. - Thanks to all of you  for attending this session.  I'm Helen Gill, Program Manger in ITO

- I'd like tell you about the new DARPA ITO program in Software Enabled Control.

- First let's roll a video for a quick overview

<video clip here>

2. - Where are we now? 

- current control systems generally trade performance for stability
- Human operators have to "close the loop" for the extremes:
(very aggressive maneuvers, severe operating environments)
- SEC will take a new look at the old assumptions: 

• Limited computing and storage resources
• Static perspective:
• Fixed-frequency schedules, computed off-line
• Monolithic control laws
• Loose coupling of switching logic, control laws
• Limited prediction (doesn't consider switching)

- New software approaches will change these

3. - New capability in three areas: 

-- Individual systems (e.g., ducted fan engine) 
-- Subsystem control  (e.g., sensors,  fire control, vehicle management)
--Cooperative control of  separate systems  (e.g., UAVs, close formation flight, coupled 
dynamics)

4. - Program Approach: 

- Raise the  bar for control performance
- Exploit computation and memory
- Better on-line use of dynamic information: control, coordination
- Exploit/Combine computer science and control theory

5. - As outlined in  the video:  Research in 4 areas:

--Active State Models
--Coordinated Multi-Modal Control
--On-Line Control Customization
--Open Control Platform

-(brief overview)

6. -The active models task:  On-Line models for large, complex systems, very large state 
and mode spaces

- Illustrate use in prediction:  mode transition - current control law becomes unsuitable 
(vulnerability; accidents)
- Blue - Normal evolution of system state (within Mode)
- Black - Environmental event that must be countered
- Red - Mode must change to respond to environmental challenge (or damage)
- Model support is needed for fast system id and prediction of viable future modes  
- Current:  Model predictive control - continuous dynamics models for control law 
adaptation
- (Supports disturbance rejection only within a mode (same control law)
- Need! Prediction:  Continuous evolution of state space + logical evolution (mode-
switching alternatives) through mode space
- Sec:  Models - bridge logical and continuous for (e.g.,)  active control of mode transition 
during severe disturbance

7. - Active state models:  data access + PREDICTION services

- Challenge 1:  combine analytic, first principles / engineering models, empirical/ on-line 
data capture, numeric/symbolic constraints
-  Challenge 2: combine models of multiple effects (system, envt) - "multi-models"  (e.g., 
weather, ground effects)
- Challenge 3: provide systems support for real-time prediction for combined models 
(types:  hybrid + BDD discrete prediction, multiscale,  content uncertainty)
- opportunities: e.g., use A.S.M. prediction to accommodate sensor  drop-out  (e.g., 
transient, vehicle structure masks the sensor)
- SEC will provide technology for predictive active state models 

8. - SECOD MAJOR SEC research area:   coordinated multi-modal control

- Problem:  need coordination of major system operational modes, with subsystem modes
- e.g., diagram:  COBRA maneuver (evasive, high AoA, lose airspeed)   
- Pullout (terrain, close-in or threat need evasive / collision avoidance)
- Need:  rapid, coordinated transition from aggressive maneuver to pull-out
- Different subsystem modes active during one phase, not another      
- Multi-vehicle systems, same problem  (coordinated kinematics, limit control of yaw 
angle)

9. -SEC must provide system support for coordination

- Single operational mode corresponds to set of supporting  subsystem modes
- And system mode transition corresponds to set of transitions in subsystems
- Transition must be coordinated to maintain correspondence
- Blue dotted lines - encapsulate subsystem modes needed by maneuver
- Red lines -  transitions that must be synchronized  
- SEC:  system support mechanisms (coord of MODES, TRANSITIONS),
- Also:  innovation in supporting theory hybrid stability, controllability and constraint 
mechanisms  (covering mode transition)

10. -Challenge:  provide multi-modal coordination services for both integrated subsystems, 
separate systems (e.g., multi-vehicle systems)  
- e.g., Autonomous close-formation flight:  coordinated kinematics, possibly coupled 
dynamics 
- Sudden changes in trajectory must be accommodated by other vehicles
- Challenge: local and global  stability & robustness, 
- Challenge:  joint failure detection and recovery 
- Challenge:  real-time coordination that exploits active state models
- Challenge:  capable of geographic distribution

11. -Third major research area:  on-line control customization

- Fast control recalculation:  high-performance, robust response (evasion, environment, 
damage)
- More than reparameterizing control laws ++ (Model Predictive control, Control Lyapunov 
Functions to guide control law optimization, preserve stability)
- Functional/parametric adaptation of control laws, also structural adaptation of modes, 
transitions
- Example 1:  Changed coordination structure, may require mode de-confliction
- Example 2:  Active transition control:  on-line mode blending
- Example 3:  Actively manage  varying sensor & actuator capability

12. - Many on-line control customization challenges 

- Solve these problems; enable higher performance
- Challenge: Preserve stability/recoverabililty during control adaptation  
- Major challenge: real-time adaptation
- Challenges require  integrated  solution:  computer science/systems & control
- SEC:  "on-line control customization" technology  will address control & systems issues 
in tandem.

13. -Fourth area:  the open control platform

- New systems approaches needed:  control middleware
- Diagram:  notional sketch, not a design (components that might populate an OCP)
Challenges:  implement & integrate services:
--Active model prediction
---Predictive (discrete and continuous) mode transition control
---Multi-modal System ID and Active-Model-referenced control laws
---On-line control re-design and customization 
----(e.g. Hybrid multimodal coordination, Model Predictive Control and event-triggered 
control recalculation using Control Lyapunov functions)

Systems Challenges:  
- Active management of MUTABLE CONTROL SCHEDULES (rather than fixed, 
statically-scheduled control loop designs)
---Model-referenced sensor/actuator resource management

14. -SEC:  software platform for open real-time embedded control systems

- Goals
--Simplify control software (both math and logic)
--Support advanced control methods that exploit on-line computation and on-line data for 
significantly higher-performance control
--Enable ongoing research and industrial adoption of advanced control technology

15. -Programmatics - evaluation & experimentation

--Baseline:  Improved stability maintenance, performance  in operation of single vehicle 
(e.g., university lab VTOL Rotorcraft)
--Challenge:  Cooperative operation of systems with coupled dynamics (e.g. joint airlift, 
simple lift mode coordination)
--Goal:  Increase envelope of operation for cooperative (including physically coupled) 
system control

16. -Transition path: possible application to performance improvements for high-
performance vehicles such as this experimental X-36 UAV prototype:
--Tailless
--Yaw-unstable
--High angle of attack, capability
--Vulnerable to environmental factors

- Baseline:  better regulation of flight, better takeoff and landing control, for remotely-
piloted, unstable vehicles
- Challenge:  Automate highly-reactive, extreme maneuver  &  evasion regimes, 
coordinated or adversarial encounters
- Goal:  Enable high- and extreme-performance autonomous tactical maneuvers in order to  
enhance survivability, capability  of autonomous weapons systems

17. - Roadmap:

The program has begun with a handful of exploratory"seedling" projects in place, and new 
starts underway:

--University of California Berkeley (hybrid multi-modal systems)
--California Institute of Technology (on-line control redesign)
--Oregon Graduate Institute (active multi-scale models and feedback-learning, software 
customization and integration)
--Honeywell (multi-model integration)
--Georgia Institute of Technology (System identification), with
--Boeing (Open  real-time embedded systems: Bold Stroke)

-Full program plans:
--Technology development in FY 2000-02, with baseline evaluation
--Technology integration, initial demonstration in FY03
--Challenge demonstration in FY04

-Potential BAAs in FY01 and FY02

18. -Reinvent control system development

- Key step along the road to a more autonomous battlefield:
--Advanced capability through software enabled control

Thanks for your attention - I'll look forward to talking with some of you this afternoon.