DARPA

Embedded Software:
Opportunities and Challenges

Dr. Janos Sztipanovits, DARPA/ITO

Dr. Shankar Sastry, Univ. of Calif. Berkeley
& formerly DARPA/ITO

Approved for public release.

 Embedded Software: Depth of DARPA
Impact
Evolution of Avionics Systems:
Advanced Avionics 1 GB
Radar
SYSTEM of SYSTEMS
Comm
•Platform Exploitation of
EW
Global Information
- Information Mining
Integrated Avionics 100 MB
Comm
Radar
NAV

Mission
Mission
Federated Avionics 1 MB
FEDERATED SUBSYSTEMS
Comm
Radar
NAV
•Functionally Integrated Data
Processing
Independent Avionics 64 KB -NAV/WD/Air Data Sensors
-Flight Control
DEDICATED SUBSYSTEMS
•Beam Steering Sensors
•Digital Fire Control/NAV
•Fly By Wire
•PT-PT Wiring
•Dedicated Digital Processing
•Mechanically Controlled
•Crew-Assisted Operations
Sensors/FLT Controls/
- Weapon Delivery
Displays
- Automated TF/TA
•Crew-Dominated Operation
- EW Response
- At-A-Distance
INTEGRATED SYSTEMS
Reconfiguration
•Aircraft-Wide Information
Integration
•Autonomous Vehicle
Emphasis
- Sensors/Stores/ Vehicle/
- Air & Space
Propulsion
•Air Crew/ Ground Crew
•Modular Electronics Monitoring & Management
•Massive Data Bases •Automated Functions
- Terrain, Threat
- ATR (Multi-Sensor)
•Digital Sensor Processing
- Failure Prognostics
- Sensor Fusion
- Route/ Sensor/ Weapon/
- Hyperspectral Imaging
Vehicle Coordination
•Integrated Diagnostics/
- Bistatic Sensing
System Fault Tolerance
(Air/ Space)
•System Data Security
- Threat Evasion
•Limited UAV Autonomy

1958 1950’s - 60’s 1970’s - 80’s 1990’s - 00’s 2000

J. Sztipanovits Source: AFRL 2
 Embedded Software: Breadth DARPA
of Impact
‹ DoD (from avionics to micro-robots)
– Essential source of new capabilities
– Largest, most complex systems

‹ Automotive (drive-by-wire)
– Key competitive element in the future
– Increasing interest but low risk taking

‹ Consumer Electronics (from mobile phones to

TVs)
– Problem is generally simpler
– US industry is strongly challenged

‹ Plant Automation Systems

– Limited market growth, conservative approach
J. Sztipanovits 3
 Overview DARPA

Impacts Development Environments Execution

Unique Representation Verification Synthesis and Frameworks
Challenges and Validation Generation
Languages Integrated Generative Models of Computation
System-Level Modeling Programming - RT support for hybrid MOC-s
Physicality Modeling - Hybrid modeling - Model-based - Integrated software/system
- Domain Specific - Model integration - Multi-target frameworks
- Aspect-Oriented - Meta-generators

Constrains Modeling Tools Analysis, Synthesis and Adaptive execution

- Meta- Application Generation frameworks, middleware
and programmable - New hybrid analysis - Decoupling via over-design
- Consistency - Constraint-based synthesis - Embeddable synthesis
Change management - Open tool integration frameworks and generation
- Reflection and runtime
Design Frameworks Framework adaptation
- Property guarantees by frameworks Composition - Probabilistic models
Variable - Orthogonalization by over-design - Middleware for
coordination
Structure - Customization

J. Sztipanovits 4
 Theme 1:
DARPA
Physicality
Embedded software: defines physical behavior of a
complex nonlinear device
Embedded System: a physical process with
Process dynamic, fault, noise, reliability, power, size
characteristics
R
Embedded Software: designed to meet required
D
A M mC physical characteristics
A
M D
Embedded
ROM M mC
Process

RAM Embedded
A
Software
F
Software
P
FPGA DSP
G
DSP
Primary challenge: How to design
A
software to achieve required physical
behavior?
Process

J. Sztipanovits 5
 Interactions, interactions,… DARPA

Controller Dynamics Embedded Software

Digital D/A Power
Plant App-1 App-2 App-3
and
Controller S/H Amp.
Sensors
Lib-1 Lib-2 Lib-3

A/D
Kernel/Services/Hardware

• control law • HW/SW architecture

• tolerated error, stability • Data types selection
• sampling rate • Scheduling policy,...
• limit-cycle oscillation • Numeric accuracy
• loop delay • Latency
• noise • Jitter

J. Sztipanovits 6
 Why Is this a Problem? DARPA

We have focused on functional

composition...

Subsystem A Subsystem B Subsystem C

Composability: Ability to link subsystems so that properties

established at subsystem levels hold at the system level

Subsystem D Subsystem E Subsystem F

But cross-cutting physical constraints

weaken or destroy composability
J. Sztipanovits 7
 Theme 2:
DARPA
Constraints and Change
Source of change: environment, requirements

Hard Problem: system-wide

Process constraints accumulate in software

Effects of changes need to

D
ROM mC
M
Process

A
EmbeddedEmbedded
Software Software propagated by tracking constraints
FPGA DSP

Flexibility is essentially a
Process
SYSTEM-WIDE CONSTRAINT
MANAGEMENT PROBLEM
J. Sztipanovits 8
 Theme 3: Dealing With Dynamic DARPA
Structures
A new category of systems:
Embedding + LARGE number of tightly integrated,
Distribution + spatially distributed physical and
information system components with
Coordination reconfigurable interconnection.

Why should we work on this?

• Tremendous progress in MEMS,
photonics, communication
technology: we need to build
systems now from these.
• Identified applications with very high
ROI: strong application pull
• Almost total lack of design theory
technology: the problem is
extremely hard.

J. Sztipanovits 9
 Networked Embedded Systems:
DARPA
Examples
MEMS Actuators for Vortex
Control (UCLA, CalTech)
• Number of nodes: 10^4
• Loop frequency: 1 KHz
• Coordination frequency: 10Hz
Vortex Control Using Distributed MEMS
• Geometric size: 30m
Actuators

Normal Vector of
Normal Vector of Reference Orbit
Pico Satellite Constellations
Cluster Plane
60o
(Aerospace Corporation)
Subsatellite Orbit
Earth
30o
• Number of nodes: 10^2 - 10^3
• Loop frequency: 1-2 KHz
Cluster plane Reference Cluster plane
• Coordination frequency: 1Hz
at t=1/2P Orbit Nadir Vector
at t=0
Information processing replaces
• Geometric size: 1-1000km
(Towards center of Earth)
mechanical structure
Key Integrating Element: I/O
I/OI/O
I/O
I/O
I/O
Modules
Modules
Modules
Modules
Modules
Modules
GP
GP
GPGP
GP
GP
Signal
Signal
Signal
Signal
Signal
Processor
Signal
Processor
Processor Super

Unified Processing Network

Processor
Processor
Processor
Image Photonic
Photonic
Processing
Image
Image Graphics Center

Interconnect Fabric Element (IFM)

Image
Processor
Image Backplane Graphics
Graphics
Processor
Processor Switch Graphics
Graphics
Processor
Processor Switch Graphics
Processor
Processor
Processor Processor
Processor
Fabric
Fabric Processor
Multiport
Multiport
Multiport Data
Multiport
Multiport
Memory
Multiport Data
Data
Memory
Memory Data
Data
Processor
Data
Memory
Memory Processor
Processor

For Avionics
Memory Processor
Processor
Processor

Interconnect Fabric Module (IFM) Integrated EO/IR System Photonic Port(s)

Switch Protocol COCKPIT

Switch Protocol Controller (SPC) Switch Protocol
Controller (SPC) Controller (SPC) DISPLAYS

Switch Protocol
Controller (SPC)
Switch Control
Unit (SCU)
Switch Protocol
Controller (SPC)
• Processing nodes: > 3*10^3
Avionics Bus

Sensors Converters
Switch Fabric

Integrated
Photonic

Vehicle FLIGHT,
Switch Protocol Switch Protocol VMS Bus

• Link Bandwidth: 2GHz

Management ELECTRICAL,
Transmitter (COTS)
FFE Logic (COTS)

FFE Logic (COTS)

Controller (SPC) Controller (SPC)

Receiver (COTS)

RFM System Photonic PROPULSION

AP RFM
Crossbar (VMS) Port(s)

Switch Protocol (COTS)

Crossbar Switch Protocol AP RFM
RFM
Controller (SPC) (COTS) Controller (SPC)
STORES
Switch Fabric

• Aggregate Bandwidth: >10^12

ASDN
Photonic

STORES
Switch Protocol Switch Protocol AP MGMT. SMS Bus
RFM
RFM STORES
Controller (SPC) Controller (SPC) SYSTEM
Switch Control (SMS)
Photonic
AP Port(s)
Unit (SCU) RFM
RFM
Switch Protocol Switch Protocol

• Industry Standard Protocol

Controller (SPC) Controller (SPC) STORES
Integrated RF System
Switch Protocol
Controller (SPC)
Switch Protocol
Controller (SPC)
Switch Protocol
Controller (SPC) Scaleable Photonics Processing
Unified Network
J. Sztipanovits 10
 Need: Middleware for DARPA
Coordination and Distribution
Model: Locally and globally relevant information
Model for global coordination

Reasoner/ Reasoner/Adapter: Adaptation of local structure

Adapter and parameters, coordination

Controller Controller: Discrete or hybrid control of local

physics

Embedded Embedded Information layer Embedded

Processing Processing Processing
Physics layer

Distribution:
Distribution: Coordination:
Coordination:
•• heterogeneous,
heterogeneous, simple
simple components
components •• global
global coordination
coordination of
of local
local interactions
interactions
(10^2-10^5)
(10^2-10^5) •• consistency
consistency of
of globally
globally relevant
relevant
•• changing
changing interconnection
interconnection topology
topology information
information
•• embedded
embedded synthesis
synthesis for
for dynamic
dynamic •• requirements
requirements are
are determined
determined byby locality
locality
distribution,
distribution, reconfiguration
reconfiguration of
of physics
physics
J. Sztipanovits 11
 DARPA Embedded Software DARPA
Program Suite

1. Software Enabled Control

FY99-03
2. Model-Based Integration of Embedded Software
FY00-04
3. Program Composition of Embedded Software
FY00-04
4. Networked Embedded Software Technology
FY01-05

J. Sztipanovits 12
 Future of Embedded Software DARPA

z One of the fundamentally new, tremendously

expanding areas of computing.
z (a) No matter which walk of life you choose,
you end up writing software…
(b) No matter how great computer scientist
you are - if you do embedded software –
you will not get away without learning
physics and engineering.
z Answering the challenges requires nothing
less than the re-integration of physical and
information sciences.
J. Sztipanovits 13
 Thematic Areas for ITO DARPA

Hu
l m
ica an
n am s Int C o
er m
Dy stem fac pu
Sy es ter
Internet
Strategic
Co Computing
m
Su pu

tra ion
bs tat

b s at
s
te
tra ion
S u f o rm
te al
s
In

J. Sztipanovits 14