Autonomous Systems Development at AFRL June 25, 2002 Paul Paul - - PowerPoint PPT Presentation

1

Autonomous Systems Development at AFRL

June 25, 2002

Paul Zetocha Group Lead, Intelligent Satellite Systems AFRL/VS (505) 853-4114 Paul.Zetocha@kirtland.af.mil Paul Paul Zetocha Zetocha Group Lead, Intelligent Satellite Systems Group Lead, Intelligent Satellite Systems AFRL/VS AFRL/VS (505) 853 (505) 853-

• 4114

4114 Paul. Paul.Zetocha Zetocha@ @kirtland kirtland. .af af.mil .mil

2

Distributed systems require autonomous decision-making among multiple satellites Distributed systems require autonomous decision-making among multiple satellites

Research in Smart Systems Research in Smart Systems Cluster Management Cluster Management

• Distributed processing
• Agent-based communication
• Formation flying process control
• Data fusion
• Virtual satellite command

and control

• Distributed processing
• Agent-based communication
• Formation flying process control
• Data fusion
• Virtual satellite command

and control Collision Avoidance – Goal directed behavior Collision Avoidance – Goal directed behavior Distributed Resource Allocation – Market negotiation Distributed Resource Allocation – Market negotiation Fault Detection Isolation & Resolution – State-based, rule-based, case-based, and model-based reasoning Fault Detection Isolation & Resolution – State-based, rule-based, case-based, and model-based reasoning Cluster geometry formation and maintenance Cluster geometry formation and maintenance – Flocking behavior – Flocking behavior

Cooperative Processing Cooperative Processing

– Genetic algorithms, neural networks, fuzzy logic – Genetic algorithms, neural networks, fuzzy logic

Message Center

Spacecraft Spacecraft Other Satellites Other Satellites

Message Center Message Center

Ground

Agent Message Center Agent Agent Agent Agent

Satellite Autonomy and Fault Detection & Recovery Satellite Autonomy and Fault Detection & Recovery

3

Cluster Manager Objectives

Formation Flying Process Control

• Relative positioning

to cm level

• Configurations from

100m to 5 Km

• Collision avoidance

On-Board Planning Command & Telemetry On-Board Science Processing

• Virtual Satellite Control
• Cluster level knowledge

maintained with SOH data passed through ISL

• Commanding to individual

satellites or to CM and then routed

• Consolidated telemetry
• Change detection
• Feature recognition
• Trigger to perform

reconfiguration

• Autonomous data recollect
• Optimization of science return
• Intelligent SV replanning due

to mission events

• Real-time

response

• Intelligent

sensor queuing

Fault Management

• On-board knowledge

base

• Limit Checking
• Real-time reaction
• CM Rollover
• SV mode maintenance

4

ASE Mission Scenario

Target Image with Autonomous Spacecraft Constellation Onboard Science Processing and Event Detection Onboard Replanning New Science Images Cluster Management: Constellation Reconfiguration

5

Information Fusion Information Fusion

The Key is Seamless FLOW to the DECISION MAKER

DATA INFORMATION KNOWLEDGE UNDERSTANDING

Terrain/Cultural Features Imagery Overlays Logistics Intelligence Weather Coalition Forces Situation

• Intel Sources
• Air Surveillance
• Surface

Surveillance

• Space

Surveillance

Fusion Technology Decision Maker

Decision-Specific Information

• Timely
• Consistent
• Structured
• Tailored
• High Quality
• Integrated

6

How Supercomputing can Enable Autonomy

Help transition to an era where we can accept and trust satellite autonomy

• High fidelity simulations of the space environment to assist in

the development, test, and evaluation of autonomous software

• High fidelity simulations of anomaly scenarios

7

How Supercomputing can Enable Autonomy

Enable Future Missions

• Model-based reasoning and planning systems that depict

satellite components to extremely high levels of detail and that can run anomaly resolution scenarios extremely fast

• Virtual reality / 3-D representations of satellites that would

allow an analyst to “step inside” and interact with individual components

• Conduct exhaustive searches of the possible results of
• perations such as switching to a redundant string