Original Intelligent Optimization Algorithms
Transforming nature, human procedures, Taiwanese culture, organizational development, and multi-agent collaboration into computable, verifiable, and transferable optimization methods.
Eight Core Original Algorithms
FBI (2020) — Forensic-Based Investigation
Translates evidence collection, investigation, tracking, and pursuit into multi-group information exchange and solution-update mechanisms for guided exploration and exploitation.
JSO (2021) — Jellyfish Search Optimizer
Models ocean-current drift, swarm motion, and active/passive feeding through a time-control mechanism. JSO has achieved broad citation impact and open-source adoption.
PWO (2023) — Pilgrimage Walk Optimization
Inspired by Taiwan's Mazu pilgrimage, translating divination blocks, group movement, stopover rituals, crawling beneath the palanquin, and palanquin robbing into search operators.
ED (2024) — Enterprise Development Optimization
Models enterprise development from local operation and strategic cooperation to knowledge learning and international expansion, balancing competition, collaboration, and knowledge transfer.
ATO (2024) — Arctic Tern Optimization
Uses long-distance migration, navigation, foraging, vigilance, and seasonal return to coordinate global exploration, directional correction, and local development.
AEIO (2026) — Age of Exploration-Inspired Optimization
Translates unknown-region exploration, route correction, resource development, and accumulated experience into mechanisms for discovery and escape from local optima.
SAPSO (2025) — Scientific Approach to Problem Solving Optimization
Follows problem definition, hypothesis formation, experimental validation, evidence analysis, and knowledge revision to support iterative learning and evidence-based search.
AAA / A³ (2026) — Agent Assembly Algorithm
Models role division, hierarchical collaboration, opinion integration, and adaptive decision making among specialized agents to solve complex engineering and computational problems.
Adoption in International Software Packages
| Algorithm | First published | Third-party package inclusion | Lag |
|---|---|---|---|
| FBI | 2020 | Python mealpy, pyMetaheuristic, pyVolutionary | ~2 months |
| JSO | 2021 | R tourr (CRAN), Python pyMetaheuristic, Opytimizer, Otorchmizer, FEALPy | ~2 months |
| ED | 2024 | Python heurilab (PyPI v2.3.0) | ~2 years |
JSO in R CRAN
Added to the R package tourr in January 2024 and shipped with CRAN release 1.2.4 in February 2025 as the public function search_jellyfish(); evaluated in a dedicated study by UT Austin and Monash researchers in the Journal of Computational and Graphical Statistics (2026).
FBI in MEALPY
Independently implemented by the open-source community in June 2020 and included in MEALPY — the largest Python metaheuristic library (233 algorithms) — under the Human category as module FBIO, providing both the original and a third-party improved class.
ED in HeuriLab
Included in the Human／Social category from PyPI release v2.3.0 (29 July 2026), alongside classics such as TLBO, Jaya and Harmony Search. The library also ships a structural-novelty audit against 19 metaphor-free criteria.
Extensions by independent research groups
JSO family
More than 15 independent groups with no co-authorship ties have developed MJS, EJS, Bin_AJS, MOQRJFS and ONBJS variants, concentrated in energy and power systems.
FBI family
Professor Ali Kaveh's group developed the enhanced EFBI, published in Computers & Structures (2021) and as a Springer book chapter (2022); Modified FBI, Self-Adaptive FBI and Robust FBI followed from groups in Türkiye, Egypt and Vietnam.
ED family
Three independent groups released improved versions — MSEDO, LMEDO and CAED — within roughly a year of publication, in Scientific Reports and Biomimetics.
Verified August 2026 against each package's public source code and its PyPI／CRAN release record.
Methodological Extensions
Fuzzy Adaptive
Dynamic adjustment of exploration and exploitation according to search feedback, including FAJSO, FAFBI, and FAATO.
Multi-objective
Pareto-front construction, external archives, and leader selection in MOJS, MOFBI, MOPWO, MOED, MOAEIO, and MOSAPSO.
Hybrid and Application-Specific
Weighted features, machine/deep learning, lightweight designs, physical constraints, and MCDM for deployable engineering-intelligence systems.
Emerging Directions in 2026
Building on the established algorithm family, PiM Lab continues work on fuzzy-adaptive Arctic Tern optimization, collective-memory evolution, black-soldier-fly-inspired intelligence, agent collaboration, and multi-objective structural design. Published methods and work in progress are displayed separately to preserve an accurate cumulative record.