【Abstract】Supply chain efficiency is the core of enterprise competitiveness, among which demand forecasting, cargo location optimization and picking path optimization are particularly critical. This article outlines the optimization plan implemented by Jiuzhang Intelligent in the warehouse management of an electrical appliance company. Demand forecasting technology uses data-driven models, such as ARIMA, to improve forecast accuracy and support inventory and market response decisions. The ALNS algorithm is used for cargo location optimization to reduce picker movement and improve picking efficiency. Picking path optimization combines OR-Tools and A* algorithms to dynamically plan efficient paths and visualize the warehouse layout through the Bokeh library. These three modules improve data consistency and system responsiveness through integration. The application of optimization algorithms not only improves operational accuracy, but also reduces human errors, bringing double improvements in cost-effectiveness and customer satisfaction to enterprises.
[Background Introduction] In today's rapidly changing business environment, the efficiency of supply chain management directly affects the competitiveness of enterprises. Demand forecasting, cargo location optimization and picking path optimization are crucial links in the supply chain, and together they form the core of enterprise warehouse management. Accurate forecasts, reasonable cargo space layout and efficient picking paths can not only reduce costs and improve operational efficiency, but also enhance the company's ability to respond to market changes. The following is a brief introduction to the plan and effects proposed by Jiuzhang Intelligent in an electrical appliance company for three core modules.
01 The first line of defense of the supply chain: demand forecasting technology
Demand forecasting is the focus of corporate gaming and has a profound impact on the company's operational efficiency, cost control and market competitiveness. Although demand forecasts are always "inaccurate", we mustReduce “head-beating” predictions, using data analysis and prediction models to make highly accurate predictions, providing decision-making basis for enterprises' inventory management, cost control, production planning, market response, etc.
In the inventory management forecast optimization project of the cooperative enterprise,The system will update data on a rolling basis every day to ensure the efficiency and accuracy of inventory management.. The specific process is as follows:
-
Customers send daily requests for pick order data, which is stored in the production database;
-
The production database sends the latest picking order data to the data cleaning module for cleaning every day;
-
The cleaned data is sent to the algorithm platform together with the logistics historical demand data;
-
The algorithm platform applies three major algorithms for analysis to generate material popularity scores, material relevance community analysis results, and material future demand forecasts.
we usedThree major algorithmsMaterials were classified, combined and forecasted:
-
ARIMA model: Autoregressive integrated moving average model, a powerful time series forecasting tool that can handle various types of time series data and has high flexibility toAdjust positive model parameters according to data characteristics, adapting it to different data sets.
-
Popularity scoring algorithm: By analyzing historical demand and frequency of use, assign a popularity score to each material and reflect it in real timeMaterial demand intensity and usage trends, helpIdentify high-demand materials 。
-
Relevance calculation algorithm: Use collaborative filtering technology and association rule mining to discover co-occurrence frequencies and usage patterns among materials,Determine the correlation between materials, optimize material combination.
We conducted a 3-month backtest, and most of the prediction errors were concentrated within a small range, showing the high accuracy of the model.
Our algorithms can also be applied in scenarios such as automated inventory replenishment systems (forecast demand, intelligent replenishment), customer order management (accurate order processing, order prioritization), data-driven decision support (data analysis and visualization, report generation), anomaly detection and early warning (inventory anomaly warning, demand anomaly detection), personalized demand customization (personalized services, customized recommendations).
02 Intelligent layout of warehouse space: cargo space optimization strategy
In the current warehousing system of cooperative enterprises, the placement of goods is usually based on fixed locations or simple classification methods, lacking scientific optimization strategies. Optimizing the placement of goods in the warehouse is an important means to improve the efficiency of warehouse management.
We proposed a cargo location optimization solution based on mathematical modeling and optimization algorithms, using a re-randomized large neighborhood search algorithm, which is a type of Adaptive Large Neighborhood Search (ALNS) algorithm. It is an efficient heuristic optimization method, especially suitable for solving combinatorial optimization problems, such as vehicle routing problems (Vehicle Routing Problem, VRP), warehouse location problems, task scheduling problems, etc.
This module analyzes the popularity, demand and relevance of goods, and combines the data set inside the warehouse to provide recommended placement plans for goods in the next month to minimize the movement distance of pickers, thereby reducing picking time costs and improving picking efficiency.
The core idea of this algorithm is to dynamically adjust the size and structure of the search neighborhood during the search process to balance the relationship between exploration and exploitation, thereby increasing the possibility of finding high-quality solutions.
The following are the visual results obtained from cargo location optimization.
03 The key to improving picking efficiency: Picking path optimization algorithm
Traditional picking methods have problems such as path redundancy and time waste. We can use algorithms to design more efficient, flexible and cost-effective warehouse picking routes.
In the warehouse picking path optimization platform, the system interacts with the warehouse management system (WMS) database to obtain material and order data in the warehouse, and uses the algorithm platform to perform optimization calculations. The specific process is as follows:
-
The client submits new optimization tasks to the algorithm platform, including picking orders and material information;
-
The algorithm platform requests the required picking order data, shelf distance data, and the actual path data of the picker from the WMS database;
-
The algorithm platform performs optimization calculations, calls the path planning algorithm, and generates optimized picking paths;
Learn the breakpoint data based on the data of the planned path and the actual walking path, continuously update the map information and optimize the picking path. The following is the application process of picking path optimization.
-
Algorithm core technology
-
Intelligent path optimization: built with the optimization toolset OR-ToolsRouting model for mixed integer programming, specifically solving the problem of picking paths. Useheuristic search strategy, quickly generate high-quality initial solutions and accelerate the solution process.
-
Dynamic path adjustment: Analyze anomalies in the picking path and useA* algorithmfind outBest detour, updating the picking route in real time to adapt to the actual situation on site. Combining historical data and real-time feedback, the algorithm parameters are continuously iteratively optimized to improve the accuracy and adaptability of path planning.
-
Warehouse layout visualization: usingBokeh librarycreateInteractive warehouse floor plan, can dynamically display different data layers, such as nodes, paths, etc. By integrating interactive tools (such as floating prompts, zooming in, saving, etc.), the interactivity of the user interface and the accessibility of information are improved.
-
Detailed function introduction
(1) Intelligent path optimization:
-
Optimal path generation: The algorithm calculates and generates the shortest picking path in the warehouse to ensure that pickers can complete the picking task with the minimum walking distance.
-
Dynamic path adjustment: Adjust picking paths based on real-time data to adapt to changes in the warehouse environment or special needs.
(2) Warehouse layout visualization
-
Interactive floor plan display: The layout of the warehouse, the location of picking points and the picking path are visually displayed through an interactive two-dimensional floor plan. Users can obtain detailed information through the graphical interface.
-
Path and node highlighting: Highlight the current picking path and important nodes, such as the starting point and end point of picking, in the floor plan to provide clear path navigation.
-
Tooltips and details: When users hover or click on a node, tooltips provide detailed information about the point, such as node number, coordinates, and associated picking information.
The following is an interactive floor plan, picking point information and shelf information of the warehouse interface.
Warehouse interface interactive floor plan
Picking point information
Shelf information
(3) Dynamic adjustment of paths
-
Real-time path update: Based on the actual walking path of the picker in the warehouse, the planned path is dynamically adjusted to deal with possible changes or unforeseen obstacles on site.
-
Data-driven optimization: Use collected path and breakpoint data to continuously optimize the path planning algorithm to make it more accurate and efficient.
The following is a visual representation of path planning.
path without breakpoints
There is a breakpoint (inside the red box) path
There are two breakpoints (inside the red box) path
-
Potential application scenarios
In the tourism industry, route planning technology helps provide personalized travel recommendation routes and generate optimized travel routes by analyzing tourists' points of interest, budget, time constraints and geographical information.
-
Points of interest analysis: Identify tourists’ points of interest (high-value points in scenic spots), such as cultural heritage, natural scenery, food experiences, etc.
-
Route optimization: Design the optimal route covering the main points of interest based on the tourist's starting point, budget and time constraints.
-
Dynamic adjustments: Adjust travel routes based on real-time traffic and weather conditions.
【Conclusion】
Through an in-depth analysis of the three key modules of demand forecasting, cargo location optimization and picking path optimization, we can see their importance and interdependence in improving warehouse management efficiency. The three are integrated through data transmission. The output of demand forecast provides a basis for cargo location optimization, and the results of cargo location optimization directly affect the planning of picking paths. This integration not only improves data consistency and accuracy, it also reduces information silos and increases the responsiveness and flexibility of the entire system.
Jiuzhang Intelligent’s optimization algorithm played a key role in this process. By applying advanced algorithms, such as ARIMA model, A* algorithm, neighborhood search algorithm, etc., we can process large amounts of data more effectively, consider multiple factors, optimize problem results, and adjust strategies in real time to adapt to market changes. These algorithms not only improve the accuracy of decision-making, but also reduce human errors, bringing significant benefits to warehouse management, including reducing costs, improving efficiency, shortening response times and improving customer satisfaction.
In addition, Jiuzhang Intelligence hopes to contribute its modest efforts to the public welfare cause of rare disease treatment. We plan to launch a free bicycle path planning problem solving algorithm, which can solve 300+ demand point problems in 10 seconds. It is open source for customers with relevant needs. We encourage customers to donate 1% of the cost savings after using the optimized solution to research on rare disease treatment. Welcome to click on the original link to join the Ruiou Scientific Innovation Alliance, a little bit of light, bright enough!
The picture above is from the Beijing News
If you have needs for demand forecasting, storage location allocation, picking route planning, etc. in the warehousing process, please feel free to consult us.
Contact: Zhang Xinjie Mobile: 13311931669 Email: [email protected]
Personal WeChat
You are also welcome to join the WeChat group below to learn about Jiuzhang Intelligence’s intelligent decision-making product series.
Welcome to click on the official account to follow more intelligent decision-making case sharing
About the author:
Liu Zhen: Graduate student in operations research and cybernetics at Nankai University. Practical experience in multiple algorithm projects. PMP project management certification.
Han Xinyue: A scientific researcher who is studying hard in the field of operations research and optimization, with experience in demand forecasting and procurement planning.
Wang Kejun: Industrial engineering major, a student studying operations research and optimization, experience in three-dimensional packaging
Liu Sijia: A novice in modeling and optimization algorithms who likes outdoor sports. She has relevant experience in project planning and order batch heuristic solving.
[Jiuzhang Intelligent Decision-making Solution] 3D-BPP Visualization Algorithm Solution
[Jiuzhang Intelligent Decision-making Solution] Vehicle Path Planning (VRP)
[Jiuzhang Intelligent Decision-making Solution] Order Wave Picking
Machine-translated from the Chinese original 【久章智能决策方案】之仓储集成优化 (published 2024-06-17). WeChat original ↗