Escaping Local Optima: A Quad-Directional Hill Climbing Algorithm for Rugged Three Dimensional Landscapes

Escaping local optima remains a central challenge in navigating rugged, multi-modal optimization landscapes. Traditional methods such as Steepest Ascent Hill Climbing (SAHC) often suffer from premature convergence due to their limited one-directional search capability. In this study, a novel QuadDirectional Hill Climbing (QDHC) algorithm is proposed to enhance the local search strategy by evaluating pseudo-gradients in four cardinal directions (North, East, West, South). This multidirectional approach enables the algorithm to make more informed decisions about ascent in complex terrains. The QDHC algorithm is evaluated on a suite of synthetic 3D benchmark functions with varying degrees of ruggedness. In highly rugged terrain, QDHC achieves an improvement of 890% in the final value of the function over SAHC, while on smoother terrains it outperforms SAHC by 96%. These results demonstrate that QDHC significantly improves convergence and robustness in non-convex, high-dimensional search spaces where traditional hill climbing often fails.

• Publication year

  2025

• Venue

  2025 International Conference on Modeling, Simulation & Intelligent Computing (MoSICom)

• Publication date

  2025-12-10

• Fields of study

  Not labeled

• Identifiers
  DOI 10.1109/MoSICom67153.2025.11398330
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• A Step-Size Adaptive Hill-Climbing Algorithm for Local Search

  2021cited by this paper
• Adam: A Method for Stochastic Optimization

  2014cited by this paper
• Motion planning for steep hill climbing

  2011cited by this paper
• Analyzing the Performance of Generalized Hill Climbing Algorithms

  2004cited by this paper
• Hybrid methods using genetic algorithms for global optimization

  1996cited by this paper
• Faster Simulated Annealing

  1995cited by this paper
• Learning long-term dependencies with gradient descent is difficult

  1994cited by this paper
• Genetic Algorithms in Search Optimization and Machine Learning

  1988cited by this paper
• Optimization by Simulated Annealing

  1983cited by this paper
• An efficient method for finding the minimum of a function of several variables without calculating derivatives

  1964cited by this paper

• No citing papers are available for this paper.