PGU-SGP: A Pheno-Geno Unified Surrogate Genetic Programming For Real-life Container Terminal Truck Scheduling

TL;DR

This work tackles the computational bottleneck of fitness evaluations in data-driven genetic programming for dynamic container terminal truck scheduling by introducing PGU-SGP, a pheno-geno unified surrogate. By integrating phenotypic and genotypic characterizations through a unified PGU distance and a node-frequency GC representation, the method improves surrogate sample selection and fitness prediction. Empirical results on real-life DCTTS instances show that PGU-SGP reduces training time by ~76% while maintaining competitive performance with traditional GP; it also outperforms the PC-only surrogate in many settings and yields faster convergence due to more accurate fitness rankings and balanced selection pressure. The approach demonstrates strong potential for accelerating surrogate-assisted GP in dynamic COPs and invites future work on broader problem domains and adaptive hyper-parameter tuning.

Abstract

Data-driven genetic programming (GP) has proven highly effective in solving combinatorial optimization problems under dynamic and uncertain environments. A central challenge lies in fast fitness evaluations on large training datasets, especially for complex real-world problems involving time-consuming simulations. Surrogate models, like phenotypic characterization (PC)-based K-nearest neighbors (KNN), have been applied to reduce computational cost. However, the PC-based similarity measure is confined to behavioral characteristics, overlooking genotypic differences, which can limit surrogate quality and impair performance. To address these issues, this paper proposes a pheno-geno unified surrogate GP algorithm, PGU-SGP, integrating phenotypic and genotypic characterization (GC) to enhance surrogate sample selection and fitness prediction. A novel unified similarity metric combining PC and GC distances is proposed, along with an effective and efficient GC representation. Experimental results of a real-life vehicle scheduling problem demonstrate that PGU-SGP reduces training time by approximately 76% while achieving comparable performance to traditional GP. With the same training time, PGU-SGP significantly outperforms traditional GP and the state-of-the-art algorithm on most datasets. Additionally, PGU-SGP shows faster convergence and improved surrogate quality by maintaining accurate fitness rankings and appropriate selection pressure, further validating its effectiveness.

Figures (6)

• Figure 1: The layout of a typical container terminal.
• Figure 2: An example of GP individual for DCTTS.
• Figure 3: The flowchart of the traditional GP algorithm.
• Figure 4: The flowchart of the proposed algorithm.
• Figure 5: The flowchart of the DCTTS simulator.