Push-pull glowworm swarm optimization algorithm for multimodal functions
Shashi Barak, Varun Raj Kompella, Krishnanand N. Kaipa, Debasish Ghose · Institution of Engineering and Technology eBooks · 2018
Glowworm swarm optimization (GSO) is a well-established swarm-intelligencebased optimization technique mainly used for identifying peaks of all local optima of multimodal functions rather than just the global optima as done by most other similar algorithms. The main application of GSO is in cases where the optima are created by sources of certain signals, which interact to create these peaks in the measured signal profile. As there is a greater likelihood of the peaks being close to the individual sources, GSO can identify such sources. It has been shown that GSO can be used to implement signal source seeking behavior in multirobot systems. The name GSO comes from the specific characteristics of glowworms in nature, which is used in the algorithm, to get attracted to other glowworms which glow brighter. Several researchers have used GSO for various applications and some of them have modified GSO to obtain better performance. In this chapter, we present a detailed study of the push-pull glowworm swarm optimization algorithm, which is a variant based on an extension of the basic philosophy of attraction in GSO. Each glowworm encodes the fitness of its current location, evaluated using the given objective function, into a luciferin value that is accessible to its neighbors. A glowworm in the proposed push-pull version of the algorithm identifies two disjoint adaptive neighborhoods, containing glowworms with higher luciferin value and lower luciferin value, for deciding its direction of movement by probabilistically selecting one neighbor from each of the neighborhoods. The lower luciferin value neighbor provides the “push” and the higher luciferin one provides the “pull.” In the originally proposed basic GSO, only the pull action was considered. This exploitation of local information and strategic neighbor interaction from two adaptive neighborhoods enables the swarm of glowworms to partition into disjoint subgroups and converge to the multiple optima of a given multimodal function with a much smoother trajectory and faster convergence. Experimental results are presented to demonstrate the efficacy of the proposed algorithm.