Multi-Objective Chaos Game Optimization Outperforms Standard Algorithms in Structural Design

Computational structural engineers have introduced Multi-Objective Chaos Game Optimization, a metaheuristic algorithm designed to resolve structural engineering problems characterized by conflicting performance criteria and behavioral constraints. Evaluated across six standard benchmark truss structures ranging from 10-bar to 120-bar configurations, the algorithm simultaneously minimized structural weight and compliance while ensuring axial member stresses remained within allowable material thresholds. Across extensive performance testing against four widely established metaheuristic techniques—NSGA-II, MOALO, MOGOA, and MOAVOA—the chaos-inspired method demonstrated superior convergence behavior and Pareto front diversity. Standardized mathematical evaluation metrics and rigorous statistical tests confirmed that Multi-Objective Chaos Game Optimization establishes an effective, robust, and computationally economical approach for constrained multi-objective structural design optimization.

The ongoing development of advanced metaheuristic optimization algorithms stems from the mounting geometric and physical complexity observed across modern engineering domains. Traditional deterministic mathematical programming techniques frequently falter when addressing large-scale structural optimization problems due to high-dimensional parameter spaces, non-convex topologies, and multimodal objective landscapes. To overcome the propensity of deterministic gradient methods to become trapped within sub-optimal local basins, computational researchers have increasingly turned to iterative stochastic approaches. These frameworks leverage stochastic exploration and exploitation mechanisms to identify near-optimal solutions within acceptable computational timeframes. Nevertheless, real-world structural engineering scenarios frequently feature conflicting physical objectives where structural mass reduction directly diminishes structural stiffness, exacerbating overall compliance under external mechanical loading. Meeting modern engineering requirements necessitates customized multi-objective metaheuristics capable of producing balanced trade-off frontiers while reliably enforcing critical operational and physical constraints.

Constrained structural multi-objective optimization requires specialized algorithmic architectures that can balance convergence toward an optimal Pareto frontier with broad diversity across high-dimensional search spaces. Although evolutionary and swarm-based algorithms have seen substantial adoption in vibration mitigation and seismic design, applying chaos-game principles to structural truss optimization has historically remained unexplored. Existing algorithms often exhibit limited solution spread or struggle with constraint satisfaction when member variables increase. To overcome these limitations, the research team adopted Chaos Game Optimization as a foundational baseline due to its fractal-geometry search mechanism. In formulating Multi-Objective Chaos Game Optimization, the overarching objective is to identify a well-distributed set of non-dominated solutions forming a comprehensive Pareto front without employing artificial scalarization or subjective objective weighting schemes.

The mathematical formulation of Multi-Objective Chaos Game Optimization begins with the simultaneous formulation of multi-objective functions subjected to a bounded feasible decision space. The decision vector contains continuous cross-sectional areas of structural truss members, which dictate structural behavior under applied mechanical loads. The algorithm initializes an initial population of one hundred candidate design solutions randomly distributed across the d-dimensional search space bounded by minimum and maximum member area constraints. To track non-dominated structural configurations discovered throughout the search process, the algorithm maintains an external Pareto archive. During successive iterations, newly evaluated candidate designs are compared against existing archive members using Pareto dominance criteria. If a new candidate design dominates an existing archive entry, the inferior entry is replaced. When an incoming candidate is non-dominated and the external archive has reached its capacity limit, a hypercube-based grid selection mechanism is deployed to eliminate crowded solutions and maintain an even distribution across objective space.

The movement dynamics of Multi-Objective Chaos Game Optimization rely on constructing temporary geometric triangles that guide the spatial exploration of candidate designs. For each candidate solution in the active population, a triangle is formed using three primary reference coordinates: the candidate solution itself, a reference leader selected from the external Pareto archive through the hypercube grid mechanism, and the mean position calculated from a randomly selected group of population members. From each temporary triangle, the algorithm generates four distinct new candidate solutions using specialized geometric update equations. The first three candidate designs are generated by adjusting coordinates relative to the candidate solution, archive leader, and group mean, scaled by exploration-exploitation coefficients and simulated dice-roll integer multipliers. The fourth candidate design incorporates a localized random mutation factor applied to a randomly selected dimension, sustaining diversity without relying on complex evolutionary operators.

To validate the algorithmic framework, the researchers tested Multi-Objective Chaos Game Optimization across six benchmark truss structures comprising 10-bar, 25-bar, 37-bar, 60-bar, 72-bar, and 120-bar configurations. The optimization formulation simultaneously minimized two conflicting engineering objectives: total structural weight and structural compliance. Structural weight was calculated as the sum of member volumes multiplied by material density, reflecting the total material investment required for fabrication. Structural compliance was calculated as the transpose of the nodal displacement vector multiplied by the external loading vector obtained through finite element analysis, representing the inverse measure of structural stiffness. All benchmark trusses utilized consistent isotropic structural material properties, featuring an allowable axial stress threshold of 400 newtons per square millimeter, a material density of 7850 kilograms per cubic meter, and an elastic Young’s modulus of 200,000 newtons per square millimeter. The optimization strictly enforced the feasibility rule during archive updates, ensuring that all candidate truss designs satisfying member stress criteria unconditionally dominated infeasible designs that violated permissible stress boundaries.

Fair algorithmic benchmarking was established by maintaining uniform operational parameters across Multi-Objective Chaos Game Optimization and the four comparative algorithms: NSGA-II, MOALO, MOGOA, and MOAVOA. Each algorithm was executed with an identical population size of one hundred candidate solutions over two hundred iteration cycles, resulting in exactly twenty thousand fitness evaluations per run. Analysis of the computational time complexity revealed that the dominant per-iteration operations in Multi-Objective Chaos Game Optimization include fitness evaluation scaling at order population size multiplied by problem dimension, archive updates scaling at archive capacity multiplied by population size, and grid crowding distance operations scaling at archive capacity multiplied by the number of grid divisions. The resulting overall worst-case complexity per run was quantified as order maximum iterations multiplied by the sum of population size times problem dimension and archive capacity times grid divisions, confirming that the algorithm’s geometric formulation avoids excessive computational overhead.

Algorithm performance was quantitatively appraised using Hypervolume and Inverted Generational Distance metrics to assess convergence accuracy and Pareto frontier diversity. Hypervolume quantifies the multidimensional objective space enclosed between the obtained non-dominated front and a predefined reference point, where larger values indicate superior convergence toward the true optimum and wider solution spread. Across all six truss instances, Multi-Objective Chaos Game Optimization achieved the highest hypervolume values. In the 10-bar truss problem, the algorithm attained an average hypervolume of 2.05 times 10 to the ninth power, outperforming NSGA-II by 3.0 percent, MOAVOA by 4.6 percent, MOALO by 5.67 percent, and MOGOA by 7.33 percent. Similar superiority appeared in the 120-bar truss configuration, where Multi-Objective Chaos Game Optimization achieved a hypervolume of 7.92 times 10 to the tenth power, demonstrating an improvement of 2.2 percent over MOALO, 4.5 percent over MOAVOA, and 13.1 percent over NSGA-II.

The Inverted Generational Distance metric provided complementary insight by calculating the average Euclidean distance from points on the true reference Pareto front to the nearest obtained non-dominated solution, where lower scores reflect closer convergence and comprehensive distribution. Multi-Objective Chaos Game Optimization consistently attained the lowest Inverted Generational Distance values across all test configurations. On the 10-bar truss structure, the algorithm produced an average Inverted Generational Distance of 1024.681, reflecting a 4.52 percent improvement over MOAVOA and a 44 percent improvement over MOGOA. In the complex 120-bar problem, the algorithm achieved a score of 8204.857, surpassing MOALO by 8.1 percent, NSGA-II by 9.0 percent, MOAVOA by 23.67 percent, and MOGOA by 36.7 percent. Comparative evolution trajectories illustrated that the algorithm exhibited faster and smoother convergence curves across successive generations, confirming that its chaotic exploration dynamics and grid-selection archive reliably accelerate convergence while sustaining an evenly populated trade-off frontier.

On individual metrics, the algorithm achieved the top average Friedman rank of 1.2167 for hypervolume and 2.3 for Inverted Generational Distance. Subsequent Nemenyi post-hoc pairwise comparisons at a significance level of alpha equals 0.05 confirmed statistically significant superiority over MOGOA on both metrics and over MOALO on hypervolume. In execution runtime benchmarks over thirty runs in MATLAB R2023b, the algorithm recorded the shortest execution times, ranging from 7.18 seconds for the 10-bar truss to 37.85 seconds for the 120-bar truss, operating roughly 1.7 times faster than MOALO and four times faster than MOGOA. The authors noted study limitations including static mechanical loading, homogeneous material assumptions, and fixed archive sizes that may constrain scalability in problems with four or more objectives, proposing future investigations into dynamic loading, topology optimization, and surrogate modeling.

Subject of Research: Multi-objective metaheuristic optimization for constrained structural engineering design

Article Title: Multi-objective chaos game optimization for constrained structural design using chaos-inspired search dynamics

Article References: Pipalia, K., Mashru, N., Jhala, R., Kalita, K., Miroslav, M., & Patel, P. (2026). Multi-objective chaos game optimization for constrained structural design using chaos-inspired search dynamics. Discover Mechanical Engineering, 5(1), Article 164. https://doi.org/10.1007/s44245-026-00336-2

Image Credits: AI Generated

DOI: 10.1007/s44245-026-00336-2

Keywords: multi-objective optimization, chaos game optimization, structural design, truss structures, hypervolume, inverted generational distance, Pareto front, finite element analysis, structural compliance, weight minimization, stress constraints, NSGA-II, metaheuristic algorithms, fractal geometry, Friedman rank test, Nemenyi post-hoc test, computational complexity