260 likes | 392 Views
Putting Simple Hierarchy into Ant Foraging: Cluster-based Soft-bots. Wei Peng, Qingmai Wang, Bin Wang, and Xinghuo Yu School of Electrical and Computer Engineering RMIT University, Melbourne. Summary. Introduction Traditional Ant Foraging algorithm Cluster-based Soft-bots algorithm
E N D
Putting Simple Hierarchy into Ant Foraging: Cluster-based Soft-bots Wei Peng, Qingmai Wang, Bin Wang, and Xinghuo Yu School of Electrical and Computer Engineering RMIT University, Melbourne
Summary • Introduction • Traditional Ant Foraging algorithm • Cluster-based Soft-bots algorithm • Simulation and Result • Discussion and Future direction
Introduction • Swarm Intelligence (SI) • SI is the property of a system whereby the collective behaviors of agents interacting locally with their environment cause coherent functional global patterns to emerge • Simple behaviors of individual agents + Communication locally with environment = Complex behavior of the group • SI-based approaches fit well into application domains whereas centralized control architecture is either unavailable or too expensive to implement.
Introduction • Key concepts of SI • Stigmergy: • Stigmergy refers to a class of mechanisms that mediate animal-animal interactions. • Indirect interaction: Two individuals interact indirectly when one of them modifies the environment and the other responds to the new environment at a later time. • Self-organization • No need for any planning and central control • No global pattern or external management
Introduction • Not many resource constraints have been considered when constructing SI-based algorithms, which result in high accumulated costs • There is a need to introduce simple hierarchy to enhance the overall performance of traditional SI approach
Introduction • A cluster-based Softbots algorithm is developed to improve the overall performance of a traditional Ant Foraging algorithm in a search-and-rescue scenario • A Layer of hierarchy is introduced to allow for cohort-like regulated behaviors so as to minimize the randomized behaviors presented in traditional SI agents.
Ant Foraging Algorithm • Two phases: • Leaving the nest to search for food • Returning to the nest with food • Pheromone: • Ants deposit pheromone on the paths that they cover to mark the trail. • Evaporation and diffusion
Ant Foraging Algorithm • Two different pheromones were used • One pheromone is deposited by ants to mark trails to the food sources. • The other pheromone is released by the nest, which diffuses in the environment and creates gradient that the ants can follow to locate the nest.
Ant Foraging Algorithm • Ant Foraging Procedure ask each ant [ if not carrying food [look-for-food] else [return-to-nest] ] diffuse ask environment [evaporate]
Ant Foraging Algorithm • return-to-nest if nest? [ drop food and head out again] else [ drop some chemical and head toward the greatest value of nest-scent] • look-for-food if food > 0 [ pick up food and reduce the food source and turn around] if sense chemical [go in the direction where the chemical smell is strongest]
Cluster-based Soft-bots Algorithm • The assumption for this approach is that simple layer of hierarchy will reduce randomized costs associated with autonomous search in Ant foraging Algorithm • Soft-bots are autonomous agents that are organized in a unit called “cluster”. • Each cluster consists of a header agent and several followers.
Cluster-based Soft-bots Algorithm • A header can recruit several followers to search for the target. They form a cluster and each follower only communicates with its header. • Once a member of a cluster identifies a target group, all members of the cluster will share the information and carry targets back to the home base.
Cluster-based Soft-bots Algorithm • A header has the ability to lay emitters, which can release signals and broadcast about the location of the identified target. • Once in the signal range of an emitter, a header will lead its cluster to move towards the emitter along the gradient of the signal. • If all targets within the range of a signal emitter have been moved away, the individual who reaches the emitter and no longer finds the target will decommission the emitter
Cluster-based Soft-bots Algorithm • Cluster-based Softbots Procedure ask each header [if find target [Drop a emitter, carry the target and return to base] else [construct cluster if detect the signal [move to emitter] else [move randomly and look-for-target] ] ]
Cluster-based Soft-bots Algorithm ask each follower [if find target [carry and return to base] else [if see the emitter [move to emitter] else [look for target if follows a header [move with header] else [move randomly] ] ] ]
Simulation and Result • The task for both algorithms is identical, which is to search for food (or target groups) and bring them to the nest (or home base). • The experiment compares time and efforts consumed in the designed task in each algorithm for various number of ants and Soft-bots. • Each algorithm uses a different strategies: ants release chemicals but Soft-bots emit signal emitters to inform other members the location of the target
Discussion and Future works • ACluster-based Soft-bots algorithm has been proposed to compensate for large portions of efforts consumed in randomized search in traditional SI-based algorithm. • The Soft-bots algorithm has been demonstrated significant comparative advantages over Ant Foraging algorithm in the presented simulation experiments.
Discussion and Future works • There is a need to validate and generalize this result under all potential operational configurations and more complicate scenario for the two algorithms. • There is also space for further improvement for the performance of the Soft-bots via introducing other regulating rules • Balance the functionality of agents and the cost of implementation