Development of a new bio-inspired optimisation algorithm
Timothy Cassell · The University of Queensland · 2016
Bio-Inspired Algorithms (BIAs) are a class of metaheuristic that have proven to be effective at optimising a vast range of complex, black box function types. A new BIA is proposed that is based on a small, nocturnal gliding possum; the native Australian sugar glider (Petaurus breviceps). Sugar Glider Algorithm (SGA) imitates the leadership hierarchy and foraging behaviour of a colony of gliders. Two co-dominant males lead a colony of five to seven gliders that forage for food by gliding between trees in search for insects or tree sap. The algorithm employs concurrent local exploitation (performed by the codominant males) and global exploration (performed by the remaining gliders). The performance of SGA has been quantitatively evaluated using five mathematical test functions, which are a mix of both unimodal and multimodal domains. The results are compared against Particle Swarm Optimisation, Differential Evolution and an Evolutionary Algorithm, with SGA performance amongst the best observed. Furthermore, SGA has been tested on three constrained engineering problems; coil spring design, welded beam design and pressure vessel design. SGA exhibited strong performance against seven existing algorithms, and found multiple new minimums than previously reported in literature. The results show that SGA is competitive against a wide range of existing algorithms in a variety of search domain topologies. These findings indicate that SGA is at the forefront of BIA performance and prove it is a superior candidate for the optimisation of engineering design problems.