Architecture%20of%20Incident%20Management%20Systems.

1
Architecture of Incident Management Systems.

• Ir. R. van der Krogt
• Ir. J. Zutt

2
Contents

• Architecture
• Replanning techniques
• Simulation (Mars)

3
Architecture (I)
Planner
Creates
Plan
Plan
Plan
Plan
Execute in
Real World
4
Architecture (II)
Planner
Creates
Adapts
Plan
Plan
Plan
Replanner
Plan
Execute in
Calls
Real World
Diagnosis
Watches
5
Strategic and Tactical level
Example Plans
Planner
Replanner
Drive(truck1, Adam, Rdam) Load(truck1,cargo1) Driv
e(truck1, Rdam, Utrecht) ...
Strategic
Observations
Diagnose
Planner
Replanner
Accelerate(truck1,20) TurnDirection(truck1,
20) TurnDirection(truck1, 0) ...
Tactical
Observations
Diagnose
Real World
6
Architecture (revisited)
Planner

• Planner creates a plan for each agent (possibly
  optimized using merging).
• Diagnosis module monitors the execu-tion and
  starts the replanner when it detects faults.

Plan
Plan
Plan
Replanner
Plan
Real World
Diagnosis
7
Replanner (I)

• Is started by the diagnosis module when it
  detects a contingency.
• Uses specialized algorithms to adapt the current
  (failing) plan to one that satisfies the goals.
• Tries to make as few changes as possible to the
  plan to avoid breaking existing commitments.

8
Replanner (II)

• Add actions

9
Replanner (II)

• Add actions
• Remove actions

10
Replanner (II)

• Add actions
• Remove actions
• Replace actions

11
Adding skills to a graph
Resource of type blue

• Extending a plan with a plan fragment.

Resource oftype pink
12
Adding skills to a graph

• Extending a plan with a plan fragment.
• Find resources that are already available in the
  plan.

13
Adding skills to a graph

• Extending a plan with a plan fragment.
• Remove skills from the plan fragment that are
  obsolete.

14
Adding skills to a graph

• Extending a plan with a plan fragment.
• Link the plan fragment to the plan.

15
Adding skills to a graph

• Extending a plan with a plan fragment.
• Final result the extended plan.

16
Simulation

• Real-world ? Simulation world.
• Why do we need simulation?
• Validation of new techniques.
• Possibility to introduce faults for testing.

17
Multi-Agent Real-Time Simulator (MARS)

• Designed by TNO-TPD.
• Written in Java, interface to Matlab/Simulink.
• Multi-Agent ? future support for multiple hosts
  / distributed simulation.
• Principally two parts Base simulator
  Experiment.

18
MARS experiment (1)
Entity behavioral model
Behavior represented using (Timed) Finite State
Machines
19
MARS experiment (1)
Entity behavioral model
Behavior represented using (Timed) Finite State
Machines
Used by the mobile entities
20
MARS experiment (1)
Entity behavioral model
Behavior represented using (Timed) Finite State
Machines
Used by the mobile entities
Initial setting, simulation goals and
introducting faults
Scenario
21
MARS experiment (1)
Entity behavioral model
Behavior represented using (Timed) Finite State
Machines
Used by the mobile entities
Initial setting, simulation goals and
introducting faults
Scenario
Visual information to display
Visual Model
22
MARS experiment (2)
Strategic observations
Entity behavioral model
Replanner
Planner
Diagnosis
Tactical observations
Real World
Scenario (initial, goals, faults)
Visual Model
23
MARS experiment (3)
Strategic observations
Entity behavioral model
Replanner
Planner
Diagnosis
Tactical observations

• Simulation step
• ?t time elapses.
• Update entities.
• Visualisation.

Real World
Scenario (initial, goals, faults)
Visual Model
24
MARS demonstration

1. Taxi-cab simulator.

25
MARS demonstration

1. Taxi-cab simulator.
2. Transport Planning.

26
MARS demonstration

1. Taxi-cab simulator.
2. Transport Planning.

• Support both layers(strategic and tactical).
• Incident Management techniqueswill be applied.

27
--- The End ---
28
Behavioral models represented with Finite State
Machines