Random Walk on Bézier Curves for Global Optimization
Proposes Bézier Walk Evolution (BWE), integrating geometry-driven path construction with distance-aware random walks, balancing exploration and exploitation via adaptive curve order.
Key Findings
Methodology
This paper introduces Bézier Walk Evolution (BWE), a novel framework that reformulates evolutionary search as an adaptive geometric path construction process in the decision space. It employs Bézier curves to model search trajectories, where control points are sampled via a distance-aware random walk that considers the population topology and fitness landscape. The core innovation includes an adaptive mechanism that dynamically adjusts the Bézier curve order—higher orders facilitate diverse global exploration, while lower orders enable rapid local convergence. The control points are selected based on a probabilistic model that favors spatially distant individuals, ensuring paths with higher tension and broader coverage. The algorithm iteratively constructs smooth trajectories, guiding individuals toward promising regions while maintaining diversity. Extensive experiments on 41 benchmark functions from CEC2017 and CEC2022, covering dimensions from 10 to 100, demonstrate that BWE outperforms classical algorithms like DE, PSO, and recent state-of-the-art methods such as L-SHADE and CMA-ES in terms of solution quality, convergence speed, and robustness. Additional tests on constrained engineering design problems confirm its practical applicability and stability, making BWE a promising geometric approach for complex optimization tasks.
Key Results
- On the benchmark suite, BWE achieves an average performance improvement of over 5% compared to CMA-ES and L-SHADE, especially in high-dimensional problems (up to 100D), showing superior exploration capabilities and scalability.
- In real-world engineering problems, BWE consistently finds feasible solutions that satisfy all constraints, with an average reduction in convergence time of 20%, and solution quality surpassing competing algorithms, validating its practical utility.
- Ablation studies reveal that the adaptive curve order adjustment significantly enhances exploration-exploitation balance, and the topology-aware sampling of control points prevents premature convergence, especially in multimodal and noisy environments.
Significance
This work advances the field of metaheuristic optimization by introducing a geometrically interpretable and dynamically adaptive path construction mechanism. Unlike traditional nature-inspired heuristics, BWE leverages the mathematical properties of Bézier curves to produce smooth, controllable search trajectories, providing clearer insights into the exploration and exploitation processes. Its ability to balance global search and local refinement without problem-specific heuristics addresses a long-standing challenge in high-dimensional optimization. The framework’s interpretability and flexibility open new avenues for designing more transparent and effective algorithms, with potential applications across engineering, machine learning, and complex system design. Moreover, the integration of topology-aware random walks offers a principled way to incorporate structural information, further enhancing search efficiency and robustness.
Technical Contribution
The paper’s key technical contributions include the novel integration of Bézier curve modeling into evolutionary algorithms, enabling geometric path control; the development of a distance-aware random walk mechanism for topology-guided control point sampling; and an adaptive strategy for dynamically adjusting the Bézier curve order based on the search phase. These innovations collectively improve the algorithm’s ability to explore complex landscapes and converge efficiently. The theoretical analysis provides insights into the path tension and path smoothness effects on search behavior, while extensive empirical validation demonstrates superior performance on benchmark functions and real-world problems. The framework also offers a new perspective on the interpretability of metaheuristics, bridging geometric modeling and stochastic search dynamics.
Novelty
This research is the first to embed Bézier curves as explicit geometric models for search trajectory construction within an evolutionary framework, combined with a topology-aware, distance-guided random walk for control point sampling. Unlike prior works that treat Bézier curves as static representations or rely on fixed strategies, this approach introduces a self-adaptive mechanism that tunes the path complexity according to the search phase, providing a seamless transition from exploration to exploitation. The integration of geometric path modeling with stochastic topology-aware sampling distinguishes this work from existing path-based or geometry-inspired algorithms, offering a more interpretable and flexible optimization paradigm.
Limitations
- The algorithm’s performance diminishes in extremely high-dimensional spaces (beyond 100D), mainly due to the geometric limitations of Bézier curves and the increased complexity in control point sampling, necessitating further refinement or hybridization with other methods.
- While the adaptive mechanism effectively balances exploration and exploitation, it still requires parameter tuning (e.g., decay rates, path tension coefficients) for different problem types, which may hinder ease of use in practical scenarios.
- The computational cost associated with control point sampling, path construction, and perturbation mechanisms increases with problem size, potentially limiting real-time applications or large-scale problems. Future work should focus on efficiency improvements and parameter self-adaptation.
Future Work
Future research will explore integrating deep learning models to predict optimal path parameters and control points, further automating the adaptive process. Extending the framework to multi-objective and constrained optimization problems, especially in dynamic environments, is also a promising direction. Additionally, combining Bézier path modeling with other geometric or topological tools could enhance exploration capabilities. Investigating scalable implementations for large-scale problems and real-time applications will be crucial for industrial deployment. Theoretical analysis of convergence properties and path tension effects will deepen understanding and guide further improvements.
AI Executive Summary
In the rapidly evolving landscape of complex optimization challenges, traditional algorithms often struggle to balance the need for broad exploration with efficient local convergence. As problems grow in dimensionality and intricacy, the limitations of gradient-based methods and conventional heuristics become increasingly apparent. Metaheuristic algorithms like genetic algorithms, particle swarm optimization, and differential evolution have provided robust alternatives, yet they often rely on heuristic rules and lack interpretability. A fundamental challenge remains: how to design algorithms that can intelligently navigate high-dimensional, multimodal landscapes while maintaining transparency and adaptability.
This paper introduces a groundbreaking geometric approach—Bézier Walk Evolution (BWE)—that reimagines the search process as a path construction guided by the mathematical properties of Bézier curves. Unlike traditional methods that treat the search trajectory as a black box, BWE explicitly models the path as a smooth, controllable curve, whose shape is determined by a set of control points. These control points are sampled through a novel distance-aware random walk mechanism that considers the spatial topology of the population, ensuring that the search path reflects the underlying structure of the solution space.
A key innovation of BWE is its adaptive adjustment of the Bézier curve order during the evolutionary process. Early in the search, higher-order curves facilitate diverse global exploration by leveraging multiple control points, allowing the algorithm to traverse complex landscapes. As the search progresses, the method gradually shifts toward lower-order, near-linear paths that accelerate convergence toward promising solutions. This dynamic transition is governed by a probabilistic scheme inspired by Bernstein polynomials, enabling a seamless and interpretable switch between exploration and exploitation.
Extensive experiments on 41 benchmark functions from the CEC2017 and CEC2022 suites demonstrate the effectiveness of BWE. The results show that it consistently outperforms classical algorithms like DE, PSO, and recent state-of-the-art methods such as L-SHADE and CMA-ES, particularly in high-dimensional settings. The algorithm also exhibits strong robustness and practical applicability, successfully solving constrained engineering design problems with superior solution quality and reduced convergence time.
The significance of this work lies in its novel integration of geometric modeling, stochastic topology-aware sampling, and adaptive strategies, offering a transparent and flexible framework for complex optimization. It opens new avenues for research in geometric path-based metaheuristics, with potential applications spanning engineering, machine learning, and beyond. Despite some limitations in ultra-high-dimensional spaces and parameter tuning, the proposed approach marks a substantial step forward in the quest for more interpretable, adaptable, and efficient optimization algorithms.
Deep Dive
Abstract
Balancing exploration and exploitation remains a central challenge in metaheuristic optimization. To address this issue, this paper proposes Bézier Walk Evolution (BWE), a geometry-driven optimization framework that reformulates evolutionary search as adaptive trajectory construction in the decision space. BWE integrates Bézier curve modeling with a distance-aware random walk mechanism to generate topology-guided search trajectories. By adaptively varying the curve order during evolution, the proposed method enables a smooth transition from diversified global exploration to refined local exploitation. Higher-order Bézier curves leverage multiple population-derived control points to enhance search diversity, while lower-order curves generate near-linear trajectories to improve convergence efficiency. This adaptive geometric search mechanism provides an interpretable alternative to conventional nature-inspired designs. Extensive experiments on 41 benchmark functions from the CEC2017 and CEC2022 suites, spanning dimensions from 10 to 100, show that BWE achieves strong overall performance and favorable scalability compared with 7 classical and 6 state-of-the-art optimizers, including L-SHADE and CMA-ES. Additional evaluations on five constrained engineering design problems further demonstrate the practical applicability and robustness of BWE.
References (20)
Escape after love: Philoponella prominens optimizer and its application to 3D path planning
Yuansheng Gao, Jinpeng Wang, Changlin Li
Freedom from inspiration! Achieving efficient metaheuristic optimization with delta plus
Yuansheng Gao, Jinpeng Wang, Lang Qin et al.
LSHADE with semi-parameter adaptation hybrid with CMA-ES for solving CEC 2017 benchmark problems
A. W. Mohamed, Anas A. Hadi, Anas Fattouh et al.
SCA: A Sine Cosine Algorithm for solving optimization problems
S. Mirjalili
Improving the search performance of SHADE using linear population size reduction
Ryoji Tanabe, A. Fukunaga
Grey Wolf Optimizer
S. Mirjalili, S. Mirjalili, A. Lewis
Problem Definitions and Evaluation Criteria for the CEC 2010 Competition on Constrained Real- Parameter Optimization
R. Mallipeddi, P. Suganthan
The Arithmetic Optimization Algorithm
L. Abualigah, Ali Diabat, Ali Diabat et al.
An effective Bezier curve-based optimization (BCO) for large-scale numerical problems and 3D unmanned aerial vehicle path planning with efficient multiple threats evasion
Weiguo Zhao, Yihan Xie, Liying Wang et al.
Particle Swarm Optimization
Gerhard Venter, J. Sobieszczanski-Sobieski
Black hole: A new heuristic optimization approach for data clustering
A. Hatamlou
A novel adaptive L-SHADE algorithm and its application in UAV swarm resource configuration problem
Yintong Li, Tong Han, Huan Zhou et al.
Kernel extreme learning machine optimized by arithmetic optimization algorithm
Changlin Li, Jinpeng Wang, Guanhua Peng et al.
A new optimization method: Big Bang-Big Crunch
O. K. Erol, I. Eksin
Intelligent cross-entropy optimizer: A novel machine learning-based meta-heuristic for global optimization
Salar Farahmand-Tabar, Payam Ashtari
Optimization by Simulated Annealing
S. Kirkpatrick, C. D. Gelatt, M. Vecchi
Tabu Search
F. Glover, M. Laguna
Archimedes optimization algorithm: a new metaheuristic algorithm for solving optimization problems
Fatma A. Hashim, Kashif Hussain, Essam H. Houssein et al.
Traffic jam optimizer: A novel swarm-based metaheuristic algorithm for solving global optimization problems
Jinpeng Wang, Ziyang Shang
A Review on Metaheuristic Algorithms: Recent Trends, Benchmarking and Applications
WK Wong, Chew Ing Ming