---

Intelligent Workflow for Collection Management

A research project funded by DARPA Solicitation BAA 97-05-PKPX, Advanced Intelligence, Surveillance and Reconnaissance Management.

Dr. Pauline M. Berry

Project Web Page: http://www.ai.sri.com/~swim/

SRI Slide

Experise

• Generative Planning
• Scheduling
• Reactive Planning/Control
• Fuzzy Control and Evidential Reasoning
• Automated Deduction
• Knowledge Management
  (browsing and editing)
• Agent Architectures
• Distributed Reasoning
• Workflow Management

CIRL Slide

Experise

• Fast Scheduling
• Search
• Planning
• Tractable Reasoning
• Optimization

Overview

• Research Objectives & Goals
• Personnel
• Technological Motivation
• State of Technology Today
• Technical Approach
• Technologies Under Investigation
• Design, Development and Demonstration Plans
• Tasks
• Schedules
• Funding
• Vision within AIM

Research Objectives

• ISR Management Requirements
  • Effective management of ISR assets
  • Satisfaction of complex set of information requirements (strategic/operational)
  • Timely delivery of information
• Research Issues
  • Information Flows
    capture information flows between agents
  • Agent Capabilities
    model the individual agent (or agent clusters) capabilities
  • Reactive & Flexible Agent Tasking
    allow rapid replanning and load balancing necessary for reactive and flexible agent tasking

Research Goals

• Revolutionary approach to Workflow Management
• Dynamic tasking of AIM agents
• Reactive and flexible control of the complete AIM Process
• Shorter AIM response cycles
• Better understanding of interactions within AIM Process

Team Members and Technical Backgrounds

Dr. Pauline M. Berry (SRI-Computer Scientist)

• Voice: (650) 859 2159
• Fax: (650) 859 3735
• Email: berry@ai.sri.com
• Technical Background: Scheduling, Uncertainty, AI Multi-agent Problem Solving Approaches to Management of Information
• Systems: PRS and MPA

Dr. Brian Drabble (CIRL-Research Associate)

• Voice: (541) 346 0473
• Fax: (541) 346 0474
• Email: drabble@cirl.uoregon.edu
• Technical Background: Intelligent Planning, Constraint- Based Search, Intelligent Workflow Generation, Information Modeling
• Systems: O-Plan

Technological Motivation

AIM-Baseline

• The current processes resemble a "stove pipe" operation and offers limited capability for reaction and replaning.
• Resources and assets are divided across a number of organisations/commands which makes coordination and replanning difficult.
• "Agents" are working in isolation with little knowledge of activities and plans of others.
• Disconnection between the need to gather intelligence ISR assets capabilities.

State of Technology Today

• Workflow Management
  • distributed and hierarchical workflow processes.
  • information flow and dependancy models
• Agent Coordination
  • collaboration/competition/communication
• Generative Planning and Scheduling Methods
  • scheduling (agent tasking/load balancing)

Technological Benefits to AIM

• Improved tasking and information flows within ISR.
• Development of qualitatively different ISR plans.
• Robustness against failure
  • resource balancing supports dynamic reallocation of assets to requests
  • here is an ISR plan and these are the points at which it is vunerable
• Dynamic development of workflow processes tailored to the current situation and context.

Workflow Management

• simple activity sequences with little dependency information
• few explicit models of the information and authority flows
• simple triggering of activities and processes
• heavily centralized with little autonomy
• limited replan capabilities with potentially large impacts on other agents and activities

• flexible process triggering and control
• context dependent operation
• information gathered on need rather than delegation

Agent Coordination

• market mechanisms (mutual selection)
• contract net (announce-bid-award)
• multi-agent planning (central agents responsible for agent tasking)
• imposed organizational structure
• emergent behavior

• planning technologies
• emergent behavior

Scheduling (Agent Tasking/Load Balancing)

• Constructive techniques:
• Constraint Satisfaction Problem (CSP) search algorithms
• Iterative Repair Techniques:
• Genetic algorithms
• Simulated Annealing
• Tabu

• reactive
• flexible agent tasking

Technical Approach

• Information Flow Modeling
• Agent Capabilities Modeling
• Workflow Allocation Strategies
• Development/Evaluation Framework
  
  

Information Flow Modeling

• Points in the process flow at which information is:
  • Created
  • Read
  • Updated
  • Authorized
• This is used to define a CRUA (created read updated authorized) matrix
• The CRUA matrix allows complex dependency recording
  
  • Given the Information Need (IN) statement, "what is the current deployment of IRAQ surface to air threats within latitudes xxx and yyy?"
  • The dependency derivations include: surface to air missiles (SAMs) in IRAQ inventory, hand launched SAMs, radar tracking acquisitions and targeting systems, support vehicles associated with SAMs, communicatiob systems associated, sensor detection capabilities, force/troop deployments within last week, month, year,..

Process Product Attributes

• The products of the process have defined attributes, e.g.
  • review status
  • approval states
  • issue status
  • availability
  • contents level
  • ....
• Sensor Tasking - Collection Management (CM) - Agent Search parameters
  • visual signal, humint, unique sensor ID capabilities - dependency derivation leading to proposed CM tasking
• The attributes have a number of defined values
  • proposed - proposed CM tasking
  • current - CM tasking
  • released - collection available to-date in all available databases (IMINT, COMINT, ELINT, MASINT, HUMINT)
  • unreleased - awaiting receipt
  • published - availability analysis to date
  • issued - current threat analysis, operational plans

Information Sources

• From the example,
  • who generated the target list?
  • who is responsible for getting it approved?
• The list came from the "threat analysis meeting" step
• The mission plan will be approved at the outcome of the "JGAT meeting" step depending on the threat analysis resulting from the ISR collection as derived from the IN..

Information Sources (cont.)

• By understanding the attributes and their potential values it becomes possible to provide better coordination":
  • "Do you have a recommended target list?"
    - "I have the approved one from yesterday and the JGAT (Joint Guidance Apportionment and Targeting) meeting to approve today's should finish by 18:00 hrs"
  • what sensors require tasking, what alternative sources should be collected based on decomposition of IN?
  • what dynamic taskings were derived based on the IN and the initial feedback of agent tasking/collection availability, results to-date?
  • what reactive/flexible control of the constellation of all available assets is necessary to complete the IN analysis?
  • what were the derived interactions?

Agent Capabilities Modeling

• The agents have specific capabilities which allow them to be mapped to steps in the process.
• The capabilities may be:
  • dynamic and vary over time
  • related to experience
• Agents may impose penalties for "overloading" which must be modeled
• The agent capabilities should be modeled in the same terms as the process steps
• The Verb/Noun/Qualifier (VNQ) model can be used for both.

Issues involved in an agent-based architecture

• Agent Topology
  • natural topology of the ISR Domain
  • stability of the topology
• Agent Connectivity
  • the nature, scale and volatility of connections
  • use made of connections
• Interdependence amongst decomposed sub-problems
• Homogeneity or heterogeneity of agents in the process

Agenda Driven Architecture

• explicit model of tasks to be solved
• explicit model of interdependencies between tasks
• explicit representation of relationships between requirements at different levels of abstraction
• natural correspondence with AIM tasks

Workflow Allocation Strategies

• Centralized Workflow Management
  
  • Processing Agents Tasked Explicitly by Centralized Controller
  • Pros: task planning; strategic view
  • Cons: less flexible reaction, problem size
• Distributed Workflow Management
  
  • Explicit Control or Implicit Control
  • Pros: flexible reaction; manageable sub-problems
  • Cons: possible chaos or lack of overall view

Technologies Under Investigation

• Modeling Techniques: VNQ (Verb/Noun/Qualifier)
• Agent Coordination: Mobile Agents, Contract Net, Task Planning..
• Agent Tasking: Adaptive/Dynamic CSPs
• Reaction: Monitors, Anytime Algorithms

VNQ (Verb/Noun Phrase(s)/Qualifier Phrase(s))

• Developed to model workflow in ACP domainVerbs describe actions to be performed
  e.g. prioritize, publish, assign
• Nouns phrases describe products of the process on which the action is to be performed
  e.g. information needs/requirement list, target list etc.
• Qualifier phrases provide information on "how" to perform the action
  e.g. broadly, roughly

VNQs in the ISR domain

• supports
  • hierarchical planning
  • libraries of alternative workflow breakdowns
  • the modeling of agent capabilities
    e.g. An agent could refine/build Information Needs (IN) lists
• modeling of constraints in AIM domain
• allows development of an advanced triggering language

Coordination Technologies

• Explicit Workflow
  • Centralized Planning Technologies
  • Distributed Tasking (re: agent topologies)
• Implicit Workflow
  • Intelligent Agents
  • softbots
  • Intelligent Information Needs Agents
  • Combination
  • Coordination Agents

Adaptive/Dynamic CSP

• Requirements
  • Agent Tasking (availability, capability, capacity)
  • Reactive (changes in world or objectives)
  • Resource allocation in the face of incomplete plans
• Technology
  • highly dynamic CSP
  • adaptive constraint satisfaction
• Related Techniques
  • situation monitoring
  • capacity analysis

Monitors, Anytime Algorithms

• Monitors
  • JFACC-like situation Monitors
  • PRS (Procedural Reasoning System)
• Anytime Algorithms
  • maintaining consistent solution

PRS

What is PRS?

• a framework for building reactive controllers
• smoothly integrates goal- and event-based activities
  • Goal-based behavior: send a robot to target location
  • Event-based behavior: respond to unexpected obstacles
• supports distributed problem-solving through multiple, communicating agents
• metalevel reasoning for defining complex control strategies

Design Development and Demonstration

• play an active role in the development of the AIM design documents
• demonstrate technology as appropriate within the overall program objectives
• support program demonstrations
• facilitate introduction of our technology into the AIM program

SOW Tasks

• Develop an ISR Workflow Process Model
  • set of workflow process tasks
  • a workflow process model
  • an information flow model
• Design and Develop an Agent-Based Workflow Framework
• Develop and Implement Agent-based Workflow Management Algorithms
• Experimentally Evaluate Agent-Based Workflow Management
• Demonstrate Agent-based Workflow in ISR
• Reports and Documentation

Schedule (Overall Project)

Schedule (6 Month Plan)

Progress

• Knowledge Acquisition
  • Domain experts
  • Program members
  • Other sources
• Literature Surveys
  • Workflow
  • Intelligent Agents
  • Collaborative Working
  • Coordination
  • Agent Organization/Topology
  • Languages (Interlingua) for Tasking/Coordination
  • CSP/Scheduling/Dynamic CSP
  • ISR/Intelligence Gathering
• Model Development
  • Information flows, capabilities, processes
• Development of Workflow philosophy for AIM: Centralized vs De-Centralized

Funding

• Current
  $330,000
• Ceiling
  $349,160
• Expenditure to-date
  $12,7317,764 (4/18/98)

Possible Impact of Technologies within AIM

• VNQ: Process Modeling
  • considers ISR process as a whole enabling exploration of more flexible interactions between components
  • maps out shorter control cycles from IN to analysis
  • development of powerful agent interlingua reflecting information flows
• Generative Planning and Adaptive CSP algorithms
  • logistics techniques adapted to AIM process
  • dynamic agent tasking through advanced scheduling technlogy
  • local plan repair and intelligent replanning
• Agent Technologies
  • will allow exploration of workflow strategies enabling flexible and dynamic control cycles

Vision for Workflow within AIM

• shorten cycles from IN request to analysis
• view ISR as a logistics operation, i.e. the right information at the right time in the right place.
• integrate provide automatic identification of need with automatic scheduling of real time tasks, i.e.
  • we know where the tanks are but where are their fuel trucks.
  • the 315th TFW just attacked a tank column at position x, y we better send over a UAV to give them BDA.
• plan and repair locally without the need to propagate through the network
• automatically identify limiting factors in plans and target ISR assets appropriately. This leads to plan robustness and the ability to replan

SWIM Project Page
AI Center Home Page
SRI International Home Page