The first DOCS "Liaohe Cup" Data-driven Complex System Optimization Algorithm Competition and the 7th International Conference on Data-Driven Optimization of Complex Systems concluded from August 19 to 21, 2025 in Taiyuan, Shanxi. The source code of the solver of some of the award-winning teams on the Synchronous Open Path Planning (VRPTW) track is now available. Interested readers, universities and industry peers are welcome to use it for verification, research and teaching.
At present, the intelligent scheduling algorithm developed by Jiuzhang Intelligence has been successfully used in aviation logistics and fresh food distribution. Partners with overseas market resources are welcome to contact us.
01 · Path Planning Track Winning Team
The first "Liaohe Cup" data-driven complex system optimization algorithm competition has the core concept of "building a communication platform, discovering outstanding talents, stimulating innovative thinking, and solving real problems" and is committed to promoting the cutting-edge development of data-driven complex system optimization. This competition has two tracks: the flexible flow workshop dynamic scheduling problem and the vehicle path planning problem with time windows. The following is an overview of the winning teams and their algorithms in the path planning (VRPTW) track. Some teams have opened the source code of their solvers.
1 · Rotating growth optimization algorithm based on variable angle interaction (Second Prize) Author: Wang Wei Instructor: Shen Yindong
Affiliation: School of Artificial Intelligence and Automation, Huazhong University of Science and Technology
Overview:VIRGO is a universal solution for VRPTW problems of any scale, aiming to break through the limitations of traditional algorithms on problem scale. The algorithm uses clustering decomposition to generate an initial solution, and uses angle rotation scanning to iteratively decompose and improve the current solution, in which the sub-problem sectors change adaptively and force adjacent sectors to overlap to avoid falling into local optimality.
2 · Enhanced version of PyVRP algorithm (second prize)
Team members: Ding Hang, Li Hangjiang, Chen Baisong, Dong Guangcheng Instructor: Luo Mao Affiliation: Hubei University of Technology
Source code:https://github.com/FSQH-dh/EPyVRP
Overview:This algorithm is based on the PyVRP framework, combines the Multi-Armed Bandit (MAB) to improve parent selection, and uses the Island Model for parallel solving to improve the solving efficiency and the quality of the solution. In order to ensure that the number of vehicles used is minimized while satisfying the capacity and time window constraints, this algorithm is divided into two stages: finding the minimum vehicle and optimizing the distance.
3 · Parallel multi-strategy subproblem decomposition HGS solver (third prize) Team members: Lu Haoze, Zheng Zubin Instructor: Liu Shengcai
Affiliation: Southern University of Science and Technology
Source code:https://github.com/0SliverBullet/DOCS2025-VRPTW
Overview:The main framework of the parallel multi-strategy subproblem decomposition HGS solver is Hybrid Genetic Search (HGS), which is a metaheuristic algorithm that is robust in combinatorial optimization problems. However, when the traditional hybrid genetic search algorithm deals with large-scale VRPTW problems, it is easy to fall into local optimality and the later search efficiency is reduced.
4·Two-stage hybrid search algorithm (third prize)
Team members: Chen Shiguang, Meng Xiyao, Yang Qilong, Yuan Rongrong Instructor: Ding Junwen
Affiliation: College of Computer Science and Technology, Huazhong University of Science and Technology
Source code:https://github.com/7YZ7/LHB-VRPTW
Overview:The team proposed an efficient heuristic algorithm to solve the vehicle routing problem with time windows. In this Liaohe Cup competition, the first optimization goal is to reduce the total number of vehicles and find the shortest path solution with the smallest number of vehicles. Therefore, this algorithm is mainly divided into two different stages. Among them, the first stage uses customer penalty and ejection pool (EP) to reduce the number of paths, and applies route weighting technology in the solution process. In the second stage, a hybrid genetic search is performed based on the minimum number of paths obtained in the first stage, and three crossover operators, namely sequential crossover algorithm (OX), edge combination crossover method (EAX) and selective route exchange (SREX), are introduced. These three crossover operators are used to perform hybrid genetics and perform local searches among the generated offspring individuals, thereby reducing the total path length of vehicles while fixing the number of vehicles.
5 · Hybrid Genetic Algorithm (Third Prize) Author: Li Yang Instructor: Shen Yindong Affiliation: Huazhong University of Science and Technology
Overview:This is a hybrid genetic algorithm (HGS, HybridGeneticalgorithm incorporated local Search) with deep integrated local search, designed to efficiently solve the vehicle routing problem with time windows (CVRPTW).
6 · Algorithm based on pyvrp framework (recursive solution for decreasing number of vehicles) (Excellence Award)
-
Captain:Zhou Hao
-
Overview:This algorithm is implemented based on the pyvrp library and designs a systematic solution process for the vehicle routing problem. The core goal is to minimize the number of vehicles used and optimize the total driving distance while satisfying constraints (such as vehicle capacity, time window, etc.).
7 · Vehicle route optimization system based on hybrid intelligent algorithm (Excellence Award)
-
Captain:Du Yue
-
Overview:This system is a vehicle routing problem (VRPTW) optimization system based on hybrid intelligent algorithms, integrating three advanced algorithms: distribution estimation algorithm (EDA), reinforcement learning (RL) and variational inference (VI).
The documents and codes of some of the winning teams above have been open sourced. You are welcome to pick them up: https://pan.baidu.com/s/138q_3aBXKidwm6x2kXLYKA?pwd=2025 (extraction code: 2025)
02 · DOCS competition results
Overall ranking of DOCS competition path planning
Final winning team (Level Award)
03 · Resource promotion
The intelligent scheduling algorithm developed by Jiuzhang Intelligence has been successfully used in aviation logistics and fresh food distribution. Currently, Jiuzhang is cooperating with Panasonic to expand SaaS services, including product forecasting, cargo location recommendations, intelligent movement routes, path planning, and three-dimensional packaging and other functions.Partners with overseas market resources are welcome to contact us.
Welcome to scan the QR code below to learn about the introduction of Jiuzhang intelligent products and capabilities and the introduction of the intelligent decision-making series.
Product and capability introduction
Intelligent decision-making series
Machine-translated from the Chinese original 🥳DOCS2025第一届“辽河杯”算法竞赛路径规划赛道获奖成果分享！ (published 2025-09-21). WeChat original ↗