Multi-objective Human-computer Co-operative Co-evolutionary Method Based on Non-dominated Sorting Strategy

Junzhou Huo, Jing Chen, Zhou Jianjun, Wu HanyangThe Open Electrical & Electronic Engineering Journal 2014 · DOI

Abstract: Based on the human-computer cooperation ideas, a Human-Computer Multi-Objective Cooperative Co-Evolutionary Method (HCMCCM) is developed to solve the complex engineering layout problem, in which the multiobjective optimization idea is integrated to avoid the "flooding" phenomenon that occurs during the combination of the artificial solutions and the algorithm solutions. In the proposed HCMCCM, the artificial solutions expressed by unified encoding strings are incorporated together with the algorithms solutions to create new cooperative solutions based on the non-dominated sorting strategies. This kind of cooperation can make the artificial solutions and the algorithm solutions on an equal basis and integrate the artificial individual with the individual algorithms into a multi-objective trade-off. The numerical simulation results of the satellite layout problem show that the proposed method can combine the artificial solutions and the algorithm solutions effectively and provide a Pareto solution set for engineers to choose from.

Introduction: The Challenge of Engineering Layouts

Designing the internal layout of complex systems, such as satellite modules, is a challenging multi-objective optimization problem. Engineers must balance competing objectives: mass distribution for stability, accessibility for maintenance, and the minimization of electromagnetic interference between components.

While computational algorithms excel at searching vast design spaces and optimizing numerical constraints, they often lack the "macro-control" and visual intuition of human designers. Conversely, human-led design is limited by the cognitive load of managing dozens of variables simultaneously. Previous attempts to merge these strengths often failed due to the "flooding" phenomenon. In these scenarios, high-quality human solutions are either "drowned" by thousands of mediocre algorithmic solutions or cause "premature convergence," where the algorithm stops exploring because a human solution is perceived as a local optimum.

The HCMCCM Framework

To address these issues, researchers developed the Human-Computer Multi-Objective Cooperative Co-Evolutionary Method (HCMCCM). This framework moves beyond simple human-replacement models, treating human expertise as an integral part of a multi-objective trade-off.

The architecture of the system (shown in Figure 1) relies on three core pillars:

  1. Unified Encoding: Both human-generated "artificial" solutions and computer-generated solutions are translated into a common string-based format. This allows the evolutionary engine to apply operators like crossover and mutation to both types of solutions interchangeably.
  2. Problem Decomposition (Co-evolution): The layout problem is broken down into sub-problems, each handled by an independent "sub-population." This cooperative co-evolutionary approach prevents the complexity of the whole system from overwhelming the search process.
  3. Non-dominated Sorting Strategy: Instead of collapsing all goals into a single fitness score, the system uses Pareto-based sorting. This ranks solutions based on their performance across all objectives. By injecting human solutions into this sorting process, they are evaluated fairly against algorithmic ones, ensuring they are preserved if they offer unique, high-performing trade-offs.
The architecture of the Multi-objective Human-computer Cooperative Co-evolutionary Model (MHCCM).

Experimental Validation: Satellite Module Layout

The researchers tested HCMCCM using a simulation of a satellite module layout involving 45 distinct objects. The goal was to place these objects on two bearing surfaces while optimizing for mass distribution (inertia), "gathering" (keeping related parts together), "disturbing" (keeping interfering parts apart), and accessibility. The method was compared against a standard Cooperative Co-Evolutionary Genetic Algorithm (CCGA) over 30 random trials.

Key Results and Insights

The HCMCCM demonstrated a significant improvement in both success rates and qualitative layout quality.

  • Success Rate: HCMCCM achieved a 95% success rate in meeting all technical constraints, compared to 85% for the CCGA.
  • Layout Quality: The layouts produced by HCMCCM showed superior "gathering" and "disturbing" scores. More importantly, the human-guided layouts were more practical.
Comparison of the best layout patterns obtained by CCGA and HCMCCM.

As seen in Figure 2, the HCMCCM layout manages the "inaccessible" and "disturbing" constraints more effectively. Human designers often have "soft" constraints in mind—such as the physical space needed for a technician's hand or a tool—that are difficult to formalize in a fitness function. The HCMCCM allows these intuitive designs to guide the algorithm toward regions of the search space that satisfy these unstated requirements.

Conclusion

The HCMCCM framework successfully integrates human intuition with algorithmic power by treating them as equals within a multi-objective optimization environment. While the method introduces a slight computational overhead (averaging ~645 seconds compared to CCGA's ~525 seconds), the resulting designs are more robust and engineer-friendly. Future work will focus on reducing human fatigue during the interactive process and scaling the system to handle layouts with hundreds of components.

Summary updated

Abstract

Based on the human-computer cooperation ideas, a Human-Computer Multi-Objective Cooperative Co-Evolutionary Method (HCMCCM) is developed to solve the complex engineering layout problem, in which the multiobjective optimization idea is integrated to avoid the "flooding" phenomenon that occurs during the combination of the artificial solutions and the algorithm solutions. In the proposed HCMCCM, the artificial solutions expressed by unified encoding strings are incorporated together with the algorithms solutions to create new cooperative solutions based on the non-dominated sorting strategies. This kind of cooperation can make the artificial solutions and the algorithm solutions on an equal basis and integrate the artificial individual with the individual algorithms into a multi-objective trade-off. The numerical simulation results of the satellite layout problem show that the proposed method can combine the artificial solutions and the algorithm solutions effectively and provide a Pareto solution set for engineers to choose from.
BibTeX
@article{huo2014multi,
  author = {Junzhou Huo and Jing Chen and Zhou Jianjun and Wu Hanyang},
  title = {{Multi-objective Human-computer Co-operative Co-evolutionary Method Based on Non-dominated Sorting Strategy}},
  journal = {The Open Electrical \& Electronic Engineering Journal},
  year = {2014},
  doi = {10.2174/1874129001408010213},
}
Figure