RRT-Based Optimizer: A Novel Metaheuristic Algorithm Based on Rapidly-Exploring Random Trees Algorithm

Abstract

Real-world optimization problems are becoming increasingly complex and require effective and versatile algorithms to provide reliable solutions. However, the no-free-lunch theorem indicates that no single optimization algorithm can solve all optimization problems accurately. Consequently, new optimization methods are required. Inspired by the search mechanism of the Rapidly-exploring Random Trees (RRT) algorithm commonly used in robot path planning, we propose a novel metaheuristic algorithm called RRT-based Optimizer (RRTO). This is the first time that the concept of the RRT algorithm has been integrated with metaheuristic algorithms. The key innovation of RRTO is its three position update strategies: adaptive step size wandering, absolute difference-based adaptive step size, and boundary-based adaptive step size. These strategies enable RRTO to efficiently explore the search space while guiding the population toward high-quality solutions. To evaluate its effectiveness, the RRTO is tested on 23 standard benchmark functions, the CEC2017 test suite, and six constrained optimization problems. In comparison with more than eight peer metaheuristic algorithms, RRTO achieves competitive results across diverse problems. Specifically, among the 35 metrics across the six constrained optimization problems, RRTO ranks first in 26 and second in 4.