As I started the day sipping tea and looking out at the trees in my backyard, I noticed flocks of birds on the horizon. The flock’s movement in unison was mesmerizing, and I watched it for a while (and recorded it for a few seconds). And that is when I decided what I would resume reading today. As you can see in the video I shot, as the swarm (or flock) changes direction, it does so in unison. No rehearsals or words have been spoken, yet a change in direction in a segment leads to the entire swarm adjusting to that change.
Change in direction, flexibility, and agility are the buzzwords in the corporate world today. But these buzzwords mostly remain talks. The primary reason behind this is the rigidity that we have built in our people, processes and systems components. This means that even though the intent may be good, the capabilities may not exist to build agility.
One challenging aspect is that we fail to understand that the core of digital transformation is not the fancy or “emerging” technologies leveraged at the periphery (including the algorithms that will be discussed in this series of articles). The core is the network of infrastructure you build. The agility in that core is what propagates agility. That core is your enterprise systems, data architecture, cloud strategy, and how they are interwoven in a network that can oscillate as business oscillates with the changing business environment.
Exactly a year ago, despite being an “analytics guy,” I insisted that organizations should make 2023 the year of data. The gist of the article was that to execute successful digital transformations, organizations must focus on building a data architecture aligned with their objectives. I am highlighting these swarm algorithms, specifically in the context of supply chains, because they can be the true answer to address some typical supply chain behaviors. However, these will not work or can not be implemented successfully if the underlying data infrastructure is not there. For example, the distributed AI approach I shared in one of my articles will be the perfect foundation to implement these algorithms. I will explain that in some examples in this article series.
This series of articles will explore the following twenty swarm optimization category algorithms. I will introduce the common sense explanation of what the algorithms aim to achieve, and then we will explore areas within supply chains where it can be leveraged.
- Ant colony algorithm
- Artificial bee colony algorithm
- Bacterial foraging optimization
- Bat optimization
- Cat swarm optimization
- Chicken swarm op[timization
- Cockroach swarm optimization
- Crow search algorithm
- Cuckoo search algorithm
- Elephant herding optimization
- Firefly algorithm
- Grasshopper optimization
- Krill herd algorithm
- Grey wolf optimizer
- Monarch butterfly algorithm
- Social spider optimization
Let us start with an algorithm that is already leveraged in some forms in various vehicle route optimization heuristics.
Ant colony optimization
As mentioned, ant colony algorithms are already leveraged, specifically in VRP solutions. I wrote a few articles on these algorithms a few years ago. In simple language, how these algorithms behave is simple. As they march, ants will tend to eventually find the shortest path to the food source. With that definition, it should not be difficult to understand why they are a good fit for VRP solutions. Ant colony optimization algorithms try to mimic the behavior. But not in the true sense, yet.
If you want some technical language, these algorithms fall within the model-based search (MBS) techniques category. MBS models are leveraged to find the optimal solution for combinatorial optimization problems. MBS models are typically categorized into two categories of algorithms. In the first category, algorithms employ a specified probabilistic model without reforming the model configuration during the run. The second category comprises algorithms that reform the probabilistic model in alternating phases. The ant colony-based techniques fall into the first class. Here are some good papers to understand ACO in the context of vehicle route optimization.
So, while the algorithm is valuable for typical VRP solutions, it can also be leveraged effectively in dynamic route optimization solutions of the future. The mathematical formulation can be extrapolated/gamed to redefine the pheromone aspect. After all, ants communicate with other ants through pheromones in real-time.
In subsequent parts of this article, we will go through the algorithms listed above in that order. The second part of this article will be published on 11/28.
Designed Analytics Blog 