Feasible

Feasible

Share this post

Feasible
Feasible
📈 #19 Warehouses - Part III-: Unlock High-Quality Solutions for Order Picking Optimization Problems

📈 #19 Warehouses - Part III-: Unlock High-Quality Solutions for Order Picking Optimization Problems

Learn how to use VNS and other hybridization techniques to tackle challenging warehouse optimization problems.

Borja Menéndez's avatar
Borja Menéndez
Dec 17, 2023
∙ Paid
3

Share this post

Feasible
Feasible
📈 #19 Warehouses - Part III-: Unlock High-Quality Solutions for Order Picking Optimization Problems
Share

As Julio Iglesias would say: hey!

Although I highly doubt Julio Iglesias would say this: today, I'm going to talk to you about two very interesting topics that will delve into algorithms to provide solutions to the Order Batching Problem...

  • The most typical variants of VNS.

  • An algorithm based on VNS, but much more powerful, to provide high-quality solutions for the OBP.

Don't tell me you don't remember the OBP; I told you about it last week 😉

So today on Feasible, we're taking it a step further... Are you with me?

Let's go for it!

🧩 The Most Typical Variants of VNS

One of the best things about VNS is that it acts as a framework rather than a strict recipe to follow. This is because modifications can be made at any phase. The most common variants include:

  • Basic VNS (BVNS): The perturbation process is random, and the improvement process involves only a local search. It can explore the solution space while achieving quality solutions in a short execution time.

  • Variable Neighborhood Descent (VND): Multiple neighborhoods are deterministically explored in different local search processes, this time without a perturbation process. VND terminates its execution at a local optimum with respect to all neighborhoods, but without perturbation, it doesn't explore other areas of the solution space.

  • General VNS (GVNS): It is an enhanced BVNS where the local search process is replaced by a VND scheme. The runtime will be longer than BVNS, but the solutions will be of much higher quality than in the two previous cases. It combines the advantages of BVNS (exploring the solution space with random moves) and VND (finding a very good local optimum).

It goes without saying that GVNS is often a highly recommended option if you want to obtain very high-quality solutions.

Having said all this, I don't want it to seem like:

  1. You understand the optimization problem.

  2. You apply VNS to obtain high-quality solutions.

  3. It's smooth sailing, and then glory, and on to the next problem.

No, no, no...

Some problems are very challenging to solve, and the algorithms you build to solve them need power-ups. In optimization, this is known as hybridization: you take good ideas from one algorithm, good ideas from another algorithm, and combine their recipes to create a super algorithm to solve your problem.

🧑🏻‍💻 A Multi-Start VNS Providing High-Quality Solutions for the OBP

This is precisely what happens with order picking optimization problems in warehouses. It's challenging to find high-quality solutions with VNS alone, but alternatives exist for that purpose.

Since these problems are complex, there is no exact algorithm that efficiently solves them. Hence, various different heuristic algorithms have been proposed.

I've done it myself, let me tell you.

Keep reading with a 7-day free trial

Subscribe to Feasible to keep reading this post and get 7 days of free access to the full post archives.

Already a paid subscriber? Sign in
© 2025 Borja Menéndez
Privacy ∙ Terms ∙ Collection notice
Start writingGet the app
Substack is the home for great culture

Share