An Overview of Swarm Intelligence and Ant Colony

An Overview of Swarm Intelligence and Ant Colony Optimization Heuristics Philipp A. Djang Ph. D. Army Research Labs "Go to the ant, thou sluggard; consider her ways, and be wise: Which having no guide, overseer, or ruler, Provideth her meat in the summer, and gathereth her food in the harvest" (Proverbs vi 6 -8)

Overview n Swarm Intelligence n Ant Colony Algorithm n Solving a Traveling Salesperson Problem n Other Examples n References

Swarm Intelligence n Swarm Intelligence (SI) is the property of a system whereby the collective behaviors of (unsophisticated) agents interacting locally with their environment cause coherent functional global patterns to emerge. n SI provides a basis with which it is possible to explore collective (or distributed) problem solving without centralized control or the provision of a global model. n Leverage the power of complex adaptive systems to solve difficult non-linear stochastic problems

Swarm Intelligence n Characteristics of a swarm: n Distributed, no central control or data source; n Limited communication n No (explicit) model of the environment; n Perception of environment (sensing) n Ability to react to environment changes.

Swarm Intelligence n Social interactions (locally shared knowledge) provides the basis for unguided problem solving n The efficiency of the effort is related to but not dependent upon the degree or connectedness of the network and the number of interacting agents

Swarm Intelligence n Robust exemplars of problem-solving in Nature n Survival in stochastic hostile environment n Social interaction creates complex behaviors n Behaviors modified by dynamic environment. n Emergent behavior observed in: n Bacteria, immune system, ants, birds n And other social animals

Ants – Swarm Intelligence Example n Franks observed Lasius Niger ants, n n n n n regulation of 1 degree Celsius range; forming bridges; raiding specific areas for food; building and protecting nest; sorting brood and food items; cooperating in carrying large items; emigration of a colony; finding shortest route from nest to food source; preferentially exploiting the richest food source available. n Without Any Central Leadership or Control

Ant Colony Optimization: Introduction n First proposed by M. Dorigo, 1992 n Heuristic optimization method inspired by biological systems n Multi-agent approach for solving difficult combinatorial optimization problems n Traveling Salesman, vehicle routing, sequential ordering, graph coloring, routing in communications networks n Has become new and fruitful research area

Ant Colony Algorithms n Algorithm was inspired by observation of real ant colonies. n Ants are essentially blind, deaf and dumb. n Ants are social creatures – behavior directed to survival of colony n Q: how can ants find the short path to food sources? n Ants deposit pheromones on ground that form a trail. The trail attracts other ants.

Ant Colony Algorithms n Ant behavior is a kind of stochastic distributed optimization behavior. n Although one ant is capable of building a solution, it is the behavior of an ensemble of ants that exhibits the shortest path behavior. n The behavior is induced by indirect communication (pheromone paths) without central control.

Ant Colony Algorithms n Ants do not know the global structure of the problem - discover the network n Limited ability to sense local environment - can only “see” adjacent nodes of immediate neighborhood. n Each ant chooses an action based on variable probability n random choice n pheromone mediated

Ant Colony Algorithms n Each ant collects information about local environment; acts concurrently and independently n No direct communication: stigmergy paradigm governs information exchange n Incremental constructive approach to building solutions n High quality solutions emerge via global cooperation.

Stigmergy n Indirect communication via interaction with environment [Gassé, 59, Wilson, 75] n Sematonic stigmergy n action of agent directly related to problem solving and affects behavior of other agents. n Sign-based n action stigmergy of agent affects environment not directly related to problem solving activity.

Pheromone Trails n Species lay pheromone trails traveling from nest, to nest or possibly in both directions. n Pheromones evaporate. n Pheromones accumulate with multiple ants using path. Nest Food source

Pheromone Trails Example E E E T=1 T=0 30 ants D 15 ants d=1. 0 d=0. 5 H C d=1. 0 B A d=0. 5 D 15 ants C H 15 ants 10 ants B 15 ants 30 ants A D 20 ants C H 10 ants B 20 ants 30 ants A

Ant Colony Algorithms n Pheromone mediated “following” behavior induces the emergence of shortest paths. n Probability of choosing a branch of a path at a certain time depends on the total amount of pheromone on the branch. n The choice is proportional to the number of ants that have used the branches.

Ant Colony Algorithms n Let um and lm be the number of ants that have used the upper and lower branches. n The probability Pu(m) with which the (m+1)th ant chooses the upper branch is:

Traveling Salesperson Problem n Famous NP-Hard Optimization Problem n Given a fully connected, symmetric G(V, E) with known edge costs, find the minimum cost tour. n Artificial ants move from vertex to order to find the minimum cost tour using only pheromone mediated trails.

Traveling Salesperson Problem n The three main ideas that this ant colony algorithm has adopted from real ant colonies are: n The ants have a probabilistic preference for paths with high pheromone value n Shorter paths tend to have a higher rate of growth in pheromone value n It uses an indirect communication system through pheromone in edges

Traveling Salesperson Problem n Ants select the next vertex based on a weighted probability function based on two factors: n n The number of edges and the associated cost The trail (pheromone) left behind by other ant agents. n Each agent modifies the environment in two different ways : Local trail updating: As the ant moves between cities it updates the amount of pheromone on the edge n Global trail updating: When all ants have completed a tour the ant that found the shortest route updates the edges in its path n

Traveling Salesperson Problem n Local Updating is used to avoid very strong pheromone edges and hence increase exploration (and hopefully avoid locally optimal solutions). n The Global Updating function gives the shortest path higher reinforcement by increasing the amount of pheromone on the edges of the shortest path.

Empirical Results n Compared Ant Colony Algorithm to standard algorithms and meta-heuristic algorithms on Oliver 30 – a 30 city TSP n Standard: 2 -Opt, Lin-Kernighan, n Meta-Heuristics: Tabu Search and Simulated Annealing n Conducted 10 replications of each algorithm and provided averaged results

Comparison to Standard Algorithms n Examined Solution Quality – not speed; in general, standard algorithms were significantly faster. n Best ACO solution 420 2 -Opt L-K Near Neighbor 437 421 Far Insert 421 420 Near Insert 492 420 Space Fill 431 421 Sweep 426 421 Random 663 421

Comparison to Meta-Heuristic Algorithms n Meta-Heuristics are algorithms that can be applied to a variety of problems with a minimum of customization. n Comparing ACO to other Meta-heuristics provides a “fair market” comparison (vice TSP specific algorithms). Best Mean Std Dev ACO 420. 4 1. 3 Tabu 420. 6 1. 5 SA 422 459. 8 25. 1

Other Application Areas n Scheduling : Scheduling is a widespread problem of practical importance. n Paul Forsyth & Anthony Wren, University of Leeds Computer Science department developed a bus driver scheduling application using ant colony concepts.

Other Application Areas n Telecommunication Networks : Network routing refers to the activity of creating, maintaining and using routing tables (one for each node in the network) to determine where to direct an incoming data stream so that it can continue its travel through the network. n In telecommunications, this is an extremely difficult problem because of the constant changes in network traffic load. The Ant Colony algorithm provides adaptive advantages that can adjust to traffic load.

Other Application Areas n Vehicle Routing Problem: The VRP is similar to the TSP, but is complicated by multiple vehicles, vehicle capacity, pickup and drop off points (which can dictate vehicle packing and scheduling). n Bernd Mullenheimer, Richard Hartl and Christine Strauss developed an Ant Colony algorithm for solving the VRP

Ant Colony Algorithms: Summary n Ant Colony Algorithms mimic Real Ants n Colony of cooperating individuals n Simulated Pheromone Trail and Stigmergy n Shortest path searching with local moves n Stochastic and myopic state transition policy n Artificial ants: n Discrete state transitions n Pheromones based on solution quality n Pheromone laying is problem dependent

Interesting Reading n Alexandrov D. , Kochetov Y. Behavior of the Ant Colony n n n Algorithm for the Set Covering Problem, Proc. of Symposium. on Operations. Research. , Springer Verlag, 2000 On the MAX/MIN Ant system, Thomas Stützle, 2001. Hybrid Ant System for the Sequential Ordering Problems, Luca Gambardella, 2002. Parallelization Strategies for Ant Colony Optimization by Thomas Stützle. In Proceedings of PPSN-V, Amsterdam, Springer Verlag, LNCS 1998 Improvements on the Ant System: Introducing the MAX-MIN Ant System by Thomas Stützle. Proceedings of Artificial Neural Nets and Genetic Algorithms 1997 The Ant System Applied to the Quadratic Assignment Problem by Maniezzo, Colorni and Dorigo. Tech. Rep. IRIDIA/94 -28, Université Libre de Bruxelles 1994

Interesting Reading n Dorigo, M. , Maniezzo, V. , Colorni, A. , The Ant System: Optimization by a Colony of Cooperating Agents, IEEE Transactions on Systems, Man and Cybernetics-Part B, v 26, n 1, 1996 n Rafael S. Parpinelli and Heitor S. Lopes and Alex A. Freitas, An Ant Colony Based System for Data Mining: Applications to Medical Data, Proceedings of the Genetic and Evolutionary Computation Conference ({GECCO}-2001) n Nicolas Monmarché, Mohamed Slimane, Gilles Venturini, Ant. Class: discovery of clusters in numeric data by an hybridization of an ant colony with the kmeans algorithm, 1999

On-Line Resources n http: //www. swarm. org/ n http: //www. swarm-bots. org/ n http: //dsp. jpl. nasa. gov/members/payman/swarm/ n http: //www. engr. iupui. edu/~shi/pso. html n http: //iridia. ulb. ac. be/~mdorigo/ACO. html n http: //www. cs. technion. ac. il/~wagner/ n http: //ants. gsfc. nasa. gov/