Benchmarking heuristic search and optimisation algorithms in Matlab

Abstract

Abstract-With the proliferating development of heuristic methods, it has become challenging to choose the most suitable ones for an application at hand. This paper evaluates the performance of these algorithms available in Matlab, as it is problem dependent and parameter sensitive. Further, the paper attempts to address the challenge that there exists no satisfied benchmarks to evaluation all the algorithms at the same standard. The paper tests five heuristic algorithms in Matlab, the Nelder-Mead simplex search, the Genetic Algorithm, the Genetic Algorithm with elitism, Simulated Annealing and Particle Swarm Optimization, with four widely adopted benchmark problems. The Genetic Algorithm has an overall best performance at optimality and accuracy, while PSO has fast convergence speed when facing unimodal problem.

A Lane Detection Algorithm under Complex Scenes

“…9 and Table. 1. Comparing with the lane detection algorithm only based on Hough transform, such as [8], the advantage of this algorithm is that the detection algorithm based on Hough transform can only deal with the straight line in near field, but the curved lane in far have no way to detect. The lane detection algorithm based on the model, for example [6], depends on the accuracy of the parameters, and because of the complex calculation, it is difficult to meet the requirement of real-time.…”

Improved Antlion Optimization Algorithm for Quadratic Assignment Problem

“…where, 𝛾 and 𝛾 are lower and upper bounds of 𝛾, 𝑥 denotes the lower bound and 𝑥 denotes the upper bound of search space [59]. Optimality metric defines the relative closeness of an objective found.…”