Operational Transport Planning with Incidents Experiments with Traplas

1
Operational Transport Planning with
IncidentsExperiments with Traplas

• Jonne Zutt and Cees WitteveenFaculty of EEMCS

2
Operational Transport Planning with Incidents
Contents Problem Methods Experiments Future work

• Operational Planning for the Pickup and Delivery
  Transportation problem
• Approach
• Research questions
• Operational Planning Methods
• Experiments
• Future work

3
Problem description
Contents Problem Methods Experiments Future work

• Pickup and Delivery Transportation Problem
  Freight has to be transported from a source to a
  destination location respecting specified time
  intervals on a transport network with limited
  capacities and speeds.
• Limited capacities lead to conflicts.

4
Approach
Contents Problem Methods Experiments Future work

• Multi-agent system, where the agents are vehicle
  planners, crossroad agents, customers,
  auctioneers, etc.
• TRAPLAS based on Pamelas RTL, a discrete event
  simulation kernel vanGemund.
• Linking experimental results to theory and use
  results to discover new relations.

5
Research questions
Contents Problem Methods Experiments Future work

• What information is necessary to obtain efficient
  planning methods (collaboration vs competition)?
• What happens when varying workload, number of
  agents (scalability), incident level (normal to
  extreme circumstances, robustness)?
• What is the relation between performance and
  characteristics of the transport network?

6
Application
Contents Problem Methods Experiments Future work

• AGV terminals (ECT)
• Underground logistic system (OLS)
• Inland shipping
• Taxiway routing

7
Planning methods
Contents Problem Methods Experiments Future work

• ?Uninformed not aware of plans of other agents.
• ?Informed planning around reservations of other
  agents.
• ?Revising priority reconsidering precedences on
  crossroads.
• ?Revising routes reconsidering routes.

8
Planning methods
T1
Contents Problem Methods Experiments Future work
T2
r1
r3
r2
r0
r4
T3
r6
r5
r8
r9
r7
r11
r10
r13
r14
r12
9
Uninformed planning
Contents Problem Methods Experiments Future work
r1
r3
r2
r0
r4
r6
r5
r8
r9
r7
r11
r10
r13
r14
r12
A1
r0
r1
r2
r3
r4
5
r4
r3
r2
r1
r0
5
A2
r4
r3
r2
r1
r0
5
A3
S 15
Time ?
10
Informed planning
Contents Problem Methods Experiments Future work
r1
r3
r2
r0
r4
r6
r5
r8
r9
r7
r11
r10
r13
r14
r12
A1
r0
r1
r2
r3
r4
5
r4
r9
r8
r7
r6
r5
r0
7
A2
r4
r9
r8
r7
r6
8
r5
r0
A3
S 20
Time ?
11
Revising priorities
Contents Problem Methods Experiments Future work
r1
r3
r2
r0
r4
r6
r5
r8
r9
r7
r11
r10
r13
r14
r12
A1
r0
9
r1
r2
r3
r4
r4
r3
r2
r1
r0
5
A2
r4
r3
r2
r1
r0
6
A3
S 20
Time ?
12
Revising routes
Contents Problem Methods Experiments Future work
r1
r3
r2
r0
r4
r6
r5
r8
r9
r7
r11
r10
r13
r14
r12
A1
r0
7
r5
r6
r7
r8
r9
r4
r4
r3
r2
r1
r0
5
A2
r4
r3
r2
r1
r0
6
A3
S 18
Time ?
13
Experiments
Contents Problem Methods Experiments Future work

• 8x8 grid networks
• 32 transport agents
• Workload varies from 160 to approx. 1000
  transportation orders
• Incident level varies from normal (no incidents)
  to severe (failure probability 0.1) circumstances

14
Increasing workload
Contents Problem Methods Experiments Future work
15
Groups of agents
Contents Problem Methods Experiments Future work
Topology grid
Informed Uninformed Revising routes (order
profits)
Wlfar_Group_1024
Workload (numbr of transportation ordrs)
16
Increasing level of incidents
Contents Problem Methods Experiments Future work
17
Future work
Contents Problem Methods Experiments Future work

• Verifying theoretical results on collaboration
  and congestion games
• Add planning methods (variants)
• Experimenting with transport network topologies
• Inland shipping

18
Questions?
Contents Problem Methods Experiments Future work

• ?

19
Modeling conflicts
Contents Problem Methods Experiments Future work