Disertasi
A novel stagnation-triggered hybrid particle swarm optimization algorithm for adapting content in serious games for team building development / Hendrawan Armanto
Abstrak
Team collaboration training is an essential element of human resource development yet traditional approaches often face cost and logistical constraints. Serious games present a promising alternative because they are interactive measurable and repeatable. However a review of existing research reveals two main gaps. First on the application side most serious games for soft skills lack gameplay designs that directly address core elements of teamwork. They also rely heavily on static rule-based AI which struggles to adapt to the complex dynamics of team strategies. Second on the algorithm side widely used methods like Particle Swarm Optimization (PSO) tend to stagnate or converge too early on poor solutions mdash especially in complex multimodal environments like strategy games. This dissertation fills existing research gaps through two main contributions. First it presents the design development and validation of a serious game called Crystal Dispatch which incorporates seven essential teamwork elements into its gameplay. Second it introduces a new optimization method called Stagnation-Triggered Hybrid Particle Swarm Optimization (STH-PSO). This algorithm uses a dynamic topology-switching strategy that balances exploration and exploitation preventing early convergence and enabling continuous search improvement. Overall the study aims to develop a validated serious game framework for teamwork training and to propose a robust adaptive algorithm to address stagnation in complex changing environments. The study employed a Research and Development (R amp D) approach adhered to the ADDIE model for game development and used an experimental design for validation. The effectiveness of Crystal Dispatch was tested using a quantitative user-perception method with 32 participants from heterogeneous (varied professions) and homogeneous (university students) groups. A 70-item Likert-scale questionnaire was used validated with Corrected Item-Total Correlation and Cronbach rsquo s Alpha. STH-PSO was tested in a Multiplayer Online Battle Arena (MOBA) game setting (Dota 2) and its performance was evaluated through benchmark tests including ten standard optimization problems scalability tests and an ablation study. Statistical analysis used the Kruskal-Wallis test with Dunn rsquo s post-hoc analysis for group comparisons and the Wilcoxon Signed-Rank Test with effect size calculations for pairwise comparisons. Results showed that Crystal Dispatch effectively improved all seven teamwork components with average scores above 3.0 on a 4-point scale. The highest gains were in Individual Accountability (mean 3.49) and Role Clarification (mean 3.39) primarily driven by the game rsquo s interdependent role design. On the algorithm side STH-PSO outperformed four other PSO variants demonstrating superior solution quality consistency and lower variability. This was confirmed by statistical results including H(4) 95.23 p lt .001 for the Rastrigin function and large to medium effect sizes (r 0.67 vs. classic PSO r 0.32 vs. lbest PSO) indicating meaningful practical improvements. In conclusion this research demonstrates that combining role-based serious game design with adaptive AI algorithms such as STH-PSO can significantly enhance team collaboration training. The study contributes two primary outcomes (1) the development of a robust STH-PSO algorithm for dynamic optimization and (2) a validated serious game framework for technology-based soft skills training. These contributions address key gaps in swarm intelligence and serious game design offering a strong foundation for future intelligent adaptive training systems.