Presentation is loading. Please wait.

Presentation is loading. Please wait.

Better Ants, Better Life? Hybridization of Constraint Propagation and Ant Colony Optimization Supervisors: Bernd Meyer, Andreas Ernst Martin Held Jun 2nd,

Similar presentations


Presentation on theme: "Better Ants, Better Life? Hybridization of Constraint Propagation and Ant Colony Optimization Supervisors: Bernd Meyer, Andreas Ernst Martin Held Jun 2nd,"— Presentation transcript:

1 Better Ants, Better Life? Hybridization of Constraint Propagation and Ant Colony Optimization Supervisors: Bernd Meyer, Andreas Ernst Martin Held Jun 2nd, 2005 - interim presentation -

2 Hybridizing CP and ACO [2] Outline Probem Area Combinatorial Optimization Problems (COPs) Constrained COPs Project Focus Where do we stand Next Steps

3 Hybridizing CP and ACO [3] Problem Area finding the best solution in a discrete set of solutions Travelling Salesman Problem (TSP) Combinatorial Optimization Problems A B C D 2 4 5 23 3 7 A – C - B - D - A

4 Hybridizing CP and ACO [4] Problem Area most COPs are NP-hard solving a NP-hard problem needs exponential time BUT: near optimal solutions in reasonable time High-level strategies that guide the search for feasible solutions stochastic Combinatorial Optimization Problems Meta-Heuristics

5 Hybridizing CP and ACO [5] Problem Area Meta-heuristic inspired by real ant behaviour using pheromone trails, ants are able to find shortest paths to food sources translating this into an algorithm, it can be used to solve COPs Ant Colony Optimization (ACO) A B C D 2 4 5 23 3 7

6 Hybridizing CP and ACO [6] Problem Area ACO can generate good near optimal solutions for various COPs… Is everything is fine, or not? No, it’s not! 

7 Hybridizing CP and ACO [7] Problem Area Example: TSP with Time windows Each city has a release data and a due date hard constraints Real world problems are constrained Hard-Constraint Handling? A B C D 8 5 30 3 15 12 (10, 50) (20, 34) (30, 35) (5, 25)

8 Hybridizing CP and ACO [8] Problem Area designed to find solutions for constraint problems make use of constraint propagation automatically reduces the domain of a constrained variable E.g. the set of cities which can be chosen during a tour construction Constraint Solving Techniques Meta-heuristics not good in handling hard constraints

9 Hybridizing CP and ACO [9] Problem Area Constraint Solving Techniques are not designed for optimization ACO is not able to handle hard constraints  How to combine them? constraint propagation A B C D 8 5 30 3 18 12 (10, 50) (20, 34) (30, 35) (11, 25) Time: 0 10 18

10 Hybridizing CP and ACO [10] investigate different coupling techniques Project Focus ACO Loose Coupling Tight Coupling ACO Constraint Propagation ACO Constraint Propagation Pheromone

11 Hybridizing CP and ACO [11] Improvement of existing algorithms e.g. CPACS – a tight coupling of ACO + CP runtime drawbacks search tree pruning using e.g. bounding techniques Analysis of ACO + Stochastic Ranking (SR) SR a way of handling constraints without using constraint propagation determine its actual functional behaviour Project Focus

12 Hybridizing CP and ACO [12] Where do we stand? implemented ACO implemented ACO + Stochastic Ranking commenced statistical analysis of ACO+SR Next Steps? implementation of CPACS algorithmic improvements of CPACS coupling of [ACO+SR] + CPACS

13 Hybridizing CP and ACO [13] Thanks for your attention! Questions?


Download ppt "Better Ants, Better Life? Hybridization of Constraint Propagation and Ant Colony Optimization Supervisors: Bernd Meyer, Andreas Ernst Martin Held Jun 2nd,"

Similar presentations


Ads by Google