Modelling a Simulation-Based Decision Support System for Effects-Based Planning Farshad Moradi, Johan Schubert [email protected] [email protected].

Modelling a Simulation-Based Decision Support System for

Effects-Based Planning

Modelling a Simulation-Based Decision Support System for

Effects-Based Planning

Farshad Moradi, Johan Schubert

[email protected]

[email protected]

Farshad Moradi, Johan Schubert

[email protected]

[email protected]

OutlineOutline

Project

Simulation-Based Decision Support

EBAO/EBP

Decision Support

Operational plan, input and output interface, finding indicators

Model

Plan, activity/action/event, actor, environment, scenario

Simulation

A*-search algorithm

Conclusions and future work

Project

Simulation-Based Decision Support

EBAO/EBP

Decision Support

Operational plan, input and output interface, finding indicators

Model

Plan, activity/action/event, actor, environment, scenario

Simulation

A*-search algorithm

Conclusions and future work

ProjectProject

Background“Real-time Simulation for Supporting Effects-Based Planning”

A three-years project commissioned by Swedish Armed Forces, started in 2008

ObjectivesDesign and develop a simulation-based decision support system for supporting EBP.

Enabling a decision maker to test and evaluate a number of feasible plans against possible courses of events (through simulation) and decide which of these plans are capable of achieving the desired military end state.

Deliver results, indicating the (so far) best sequence of activities (plan option), at each point of time.

Background“Real-time Simulation for Supporting Effects-Based Planning”

A three-years project commissioned by Swedish Armed Forces, started in 2008

ObjectivesDesign and develop a simulation-based decision support system for supporting EBP.

Enabling a decision maker to test and evaluate a number of feasible plans against possible courses of events (through simulation) and decide which of these plans are capable of achieving the desired military end state.

Deliver results, indicating the (so far) best sequence of activities (plan option), at each point of time.

Simulation-Based Decision SupportSimulation-Based Decision Support

PhysicalSystemPhysicalSystem

InputData fusionOptimizer

InputData fusionOptimizer

Collect dataMeasure

What if?Experiments

Simulate

Simulate

Simulate

Simulate

Implement- Control- Decision support

Automatic Validation

OutputAnalysisOutput

Analysis

Effects-based Approach to OperationsEffects-based Approach to Operations

A process for obtaining a desired strategic outcome or “effect” on the enemy, through the synergistic, multiplicative, and cumulative application of the full range of military and non-military capabilities at the tactical, operational, and strategic levels [USJFCOM, 2005]

A process for obtaining a desired strategic outcome or “effect” on the enemy, through the synergistic, multiplicative, and cumulative application of the full range of military and non-military capabilities at the tactical, operational, and strategic levels [USJFCOM, 2005]

Effects-Based Analysis: Which effects are caused by actions?Effects-Based Analysis: Which effects are caused by actions?

(Causal)Mechanisms

Hostile Actions

Own Actions

EffectsEffect

Effect

Other Actor’sActions

Decision Support:The operational plan

Decision Support:The operational plan

Analyzing and simulating the operation plan at any time.

Analyze and simulate several alternative plans that are in the main direction of interest.

The goal is to find robust groups of plans that have similar implications.

Analyzing and simulating the operation plan at any time.

Analyze and simulate several alternative plans that are in the main direction of interest.

The goal is to find robust groups of plans that have similar implications.

Decision Support: Input InterfaceDecision Support: Input Interface

The user may select area of interestThe user may select area of interest

Decision Support: Output InterfaceDecision Support: Output Interface

Decision support is given as the most robust operational plans (left).Decision support is given as the most robust operational plans (left).

Decision Support: Output InterfaceDecision Support: Output Interface

Intelligence indicators are found by the systems (right).Intelligence indicators are found by the systems (right).

Decision Support: finding indicatorsDecision Support: finding indicators

We can support the intelligence service by finding indicators through simulations.

The hypothesis is that there are groups of plans with similar consequences.

These indicators describe the dividing line between groups of different plans.

If plans cross these lines consequences will be drastic.

We can support the intelligence service by finding indicators through simulations.

The hypothesis is that there are groups of plans with similar consequences.

These indicators describe the dividing line between groups of different plans.

If plans cross these lines consequences will be drastic.

ModelModel

Modelling based on EBAO and its concepts

PlanA sequence of activities that together lead to a desired end-state which is set by a military force

ActivityAn event initiated by own forces, which requires different types of resources in order to be executed

Event1)Initiated by our own actor (activity), 2)initiated by other actors (could be either planned or responsive), 3)spontaneous/natural events (unpredicted incidents, such as weather conditions, natural

catastrophes, an unprovoked attack or an accident)

Modelling based on EBAO and its concepts

PlanA sequence of activities that together lead to a desired end-state which is set by a military force

ActivityAn event initiated by own forces, which requires different types of resources in order to be executed

Event1)Initiated by our own actor (activity), 2)initiated by other actors (could be either planned or responsive), 3)spontaneous/natural events (unpredicted incidents, such as weather conditions, natural

catastrophes, an unprovoked attack or an accident)

ModelModelActor

An entity with resources, an action repertoire, an agenda and an internal state

Entity: group of people, who somehow have a common identity and purpose

organized such as police forces, relief agencies, well-organized militia units, and state administrative bodies

loosely coupled groups and social clusters, which are only held together by one common interest

a single individual, such as a prominent opinion maker, a political leaders or a financial potentate

Action repertoire

a set of possible actions that an entity is capable of performing, determined by its resources and knowledge

Each action has a probability of being executed, which is dynamic

ActorAn entity with resources, an action repertoire, an agenda and an internal state

Entity: group of people, who somehow have a common identity and purpose

organized such as police forces, relief agencies, well-organized militia units, and state administrative bodies

loosely coupled groups and social clusters, which are only held together by one common interest

a single individual, such as a prominent opinion maker, a political leaders or a financial potentate

Action repertoire

a set of possible actions that an entity is capable of performing, determined by its resources and knowledge

Each action has a probability of being executed, which is dynamic

ModelModel

Actor

Agenda

the plan that an actor is supposed to follow in order to achieve its goals

State

a combination of resources, such as weapon strength, no of soldiers, etc., and internal state, such as mood, solidarity, short-term agenda, etc. The states of the actors changes as a response to the activities and events, together with the probability of performing different actions

Actor

Agenda

the plan that an actor is supposed to follow in order to achieve its goals

State

a combination of resources, such as weapon strength, no of soldiers, etc., and internal state, such as mood, solidarity, short-term agenda, etc. The states of the actors changes as a response to the activities and events, together with the probability of performing different actions

ModelModel

Actor State attributes, an example:Actor State attributes, an example:Resources:

Weapon Strength firepower movement

Crew number capable of bearing arms number of sympathizers location

Economy scale stability spatial dominance

Logistical capacity to use resources optimally infrastructure propaganda channels

Soft power contacts reputation

Resources: Weapon Strength

firepower movement

Crew number capable of bearing arms number of sympathizers location

Economy scale stability spatial dominance

Logistical capacity to use resources optimally infrastructure propaganda channels

Soft power contacts reputation

Internal State:

Discontent - experienced distance to the ideal desired end state

Relationships - the degree of aversion to each of the other players

Teamwork – cohesion

Ideological conviction

Purposefulness

Cunning - wisdom

Internal State:

Discontent - experienced distance to the ideal desired end state

Relationships - the degree of aversion to each of the other players

Teamwork – cohesion

Ideological conviction

Purposefulness

Cunning - wisdom

ModelModel

Actor action repertoire, an example:Military actions:

Bomb plants

Bomb transports

Insulate and tie the opponent's resources

Regroup

Eliminate opponents positions

Secure transport corridor

Secure storage area

Secure area

Search an area

Prevent view

Sniper

Capitulation

Actor action repertoire, an example:Military actions:

Bomb plants

Bomb transports

Insulate and tie the opponent's resources

Regroup

Eliminate opponents positions

Secure transport corridor

Secure storage area

Secure area

Search an area

Prevent view

Sniper

Capitulation

ModelModelScenario

consists of participating actors, their initial state and probability distribution for different actions, environmental data, as well as the plan that is to be evaluated and an event list which consists of actions derived from the other actors’ agendas, and spontaneous/natural events

Environment

consists of various facilities and sites with symbolic value.

Functional buildings, such as hospitals, schools, housing, management centres, etc.

Transportation routes and transfer points, such as roads, bridges, pipelines, ports, airports, etc.

Utilities such as natural resources like arable land, mines, etc. and processing facilities such as power plants, factories, warehouses, etc.

Information channels such as radio and TV stations, networks, transmission masts, etc.

The symbolical sites can be geographical areas, statues or other memorials, religious buildings, etc.

Scenario

consists of participating actors, their initial state and probability distribution for different actions, environmental data, as well as the plan that is to be evaluated and an event list which consists of actions derived from the other actors’ agendas, and spontaneous/natural events

Environment

consists of various facilities and sites with symbolic value.

Functional buildings, such as hospitals, schools, housing, management centres, etc.

Transportation routes and transfer points, such as roads, bridges, pipelines, ports, airports, etc.

Utilities such as natural resources like arable land, mines, etc. and processing facilities such as power plants, factories, warehouses, etc.

Information channels such as radio and TV stations, networks, transmission masts, etc.

The symbolical sites can be geographical areas, statues or other memorials, religious buildings, etc.

SimulationSimulation

An activity An transforms the system state Sn according to Sn = f (Sn-1, An), in the time interval (tn-1, tn)

Sn is the sum of the actors’ and environment states

f (Sn-1, An) is implemented as an event-driven, stochastic simulation

Simulates interactions between our own activity, other actors’ agendas and response operations, and other external events

Monte Carlo simulations are used in order to obtain frequency functions of the entire outcome space

An activity An transforms the system state Sn according to Sn = f (Sn-1, An), in the time interval (tn-1, tn)

Sn is the sum of the actors’ and environment states

f (Sn-1, An) is implemented as an event-driven, stochastic simulation

Simulates interactions between our own activity, other actors’ agendas and response operations, and other external events

Monte Carlo simulations are used in order to obtain frequency functions of the entire outcome space

SimulationSimulationFor each round of the Monte Carlo loop:

Initialize event list with our activity A

Randomly draw the external events and add them to the event list

Randomly draw a starting state for each state parameter from resp. distribution.

For each actor:

Randomly draw the next action from the current agenda and add to the event list.

For each event in the event list as long as time is less than tn:

Environmental parameters may change (which could generate new events).

For each actor (including "our own" operator "):

– Note directly or indirectly through filtered or biased information.– Analyse the information → internal state and resources are changing.– Action repertoire is updated with new probabilities – Randomly generate the next action– Add a new action to the event list.

Save the results for each state parameter.

Create a summary of results for each state parameter in the form of a histogram, which serves as an approximation for resp. output distribution

For each round of the Monte Carlo loop:

Initialize event list with our activity A

Randomly draw the external events and add them to the event list

Randomly draw a starting state for each state parameter from resp. distribution.

For each actor:

Randomly draw the next action from the current agenda and add to the event list.

For each event in the event list as long as time is less than tn:

Environmental parameters may change (which could generate new events).

For each actor (including "our own" operator "):

– Note directly or indirectly through filtered or biased information.– Analyse the information → internal state and resources are changing.– Action repertoire is updated with new probabilities – Randomly generate the next action– Add a new action to the event list.

Save the results for each state parameter.

Create a summary of results for each state parameter in the form of a histogram, which serves as an approximation for resp. output distribution

A*-search algorithmA*-search algorithm

One of the main requirements of the simulation system is:

At any moment in time, suggest an alternative sequence of activities that best suits the decision maker’s desired end-state

Requires an algorithm that searches through the activity tree in an efficient manner so that there is always a so far ”best option” available for presentation

Neither “breadth first search” nor “depth first search” can meet this requirement. A*-search is the solution

One of the main requirements of the simulation system is:

At any moment in time, suggest an alternative sequence of activities that best suits the decision maker’s desired end-state

Requires an algorithm that searches through the activity tree in an efficient manner so that there is always a so far ”best option” available for presentation

Neither “breadth first search” nor “depth first search” can meet this requirement. A*-search is the solution

A*-search algorithmA*-search algorithm

S0 100

S11 84 S12 79 S13 103

A13 A12 A11

S0 100

S11 84 S12 79 S13 103

A13 A12 A11

S221 88 S222 71

A221 A222

S0 100

S11 84 S12 79

S13 103

A13 A12 A11

S221 88 S222 71

A221 A222

S3221 98 S3222 112 S3223 87

A3223 A3222 A3221

S0 100

S11 84 S12 79 S13 103

A13 A12 A11

S221 88 S222 71

A221 A222

S3221 98 S3222 112 S3223 87

A3223 A3222 A3221

S211 108 S212 59

A211 A212

Step 4: Activities following S11 are now simulated and S212 is the “closest” and next to simulate.

Step 2: After execution of alternative activities that follow S12, S222 is the “closest” to the target.

Step 3: From S222 all the alternative activities that are presented are executed. S11, which was calculated earlier appears to be “closest” now.

Step 1: From the initial state all available alternatives are simulated. S12 appears to be “closest” to the target.

Conclusions and future workConclusions and future workWe have designed of a simulation-based decision support methodology with which we can test operational plans as to their robustness

We have suggested a methodology that can find important indicators, towards which the intelligence service may put intelligence questions

The system is still under development hence there are no experimental results obtained so far

The current version is being tested at the moment

Future work includes testing the system with actual operational plans, a more precise actor profiles, and detailed functions for

calculating probabilities of actions in the action repertoires

We have designed of a simulation-based decision support methodology with which we can test operational plans as to their robustness

We have suggested a methodology that can find important indicators, towards which the intelligence service may put intelligence questions

The system is still under development hence there are no experimental results obtained so far

The current version is being tested at the moment

Future work includes testing the system with actual operational plans, a more precise actor profiles, and detailed functions for

calculating probabilities of actions in the action repertoires