1.  Storage Space Allocation at Marine Container Terminals Using Ant-based Control by
Omor Sharif and Nathan Huynh
Session 677:
Innovations in intermodal terminal design and operations
Sponsored by AT020, AT050, AW010

2. Overview of Storage Space Allocation Problem (SSAP)

3. Overview of SSAP Container yard is divided up into yard blocks
Each yard block holds a group of containers
Each yard block has a fixed total capacity
Storage changes with time and among blocks
Assign yard blocks to newly arriving containers

4. Objectives of SSAP Two competing objectives:
Reduce Workload Imbalance – Proper distribution of newly arriving containers amongst blocks
Minimize Container Transportation Cost – Minimize the distance traveled by ITs between berth and yard

5. Two Levels of SSAP Block level decision-
Finds the block to store a container
Objectives are ‘workload balancing’ & ‘distance minimizing’
Stack level decision-
Finds an exact stack in a block
Objective is to ‘minimize future reshuffling’

6. Why Workload Balancing?
Reduces relocation of yard cranes
Reduces turn time of vessels
Reduces average container handling time
Reduces congestion on road network
Improves efficiency of related resources

7. What is Ant-Based Control?
Falls under ‘Ant colony optimizations & algorithms’
Used to solve complex routing problems in network
Inspired by collective intelligence of ants
Can be used for shortest path and load balancing
Utilizes pheromone laying behavior of ants
Ants have a decision to make
Situation several months later

8. Ant Based Control for SSAP
A network approach to container terminal
Network support artificial ant agents
Ants move just like inbound & outbound containers
Lays pheromone as a function of distance & congestion
Probabilities in pheromone table guide ants & containers
Fig: Ant choses path according to probability

9. Shortest Path Workload Balancing
Ant age Increases with
traveled distance
Probability of selection
decrease with ant age
Workload Balancing
Ants are delayed at congested blocks
Delay is proportional to degree of congestion

10. Dynamic Interaction between Ants & Containers

11. Simulation and Experiments
A simulation was modeled in NetLogo
NetLogo - A multi-agent programmable environment

12. Parameter and Values Used in Model

13. Parameter and Values Used in Model

14. Performance Metric I - Transport Distance
Converge to shortest path in 500 time-units
Under no congestion converge to lower bound
When congestion occurs distance converge to 2% above lower bound

15. Performance Metric II - Workload Imbalance
Standard deviation gradually rises when no congestion
When congestion occurs standard deviation falls
Maintains a distribution to accommodate both objectives

16. Conclusions What’s New?
Distributed allocation to individual containers
Real-time approach requiring no advanced container arrival information
Adaptive and scalable
Synchronous instead of hierarchical approach
Potential for model integration
Future Work
Incorporate ‘stack level’ decision
Categorization of containers (Loaded, Empty etc.)

17. Thank You