Improving productivity is undoubtedly a continuous goal for container terminal operators in Colombia. The task of sustaining or improving operational efficiency along with financial margin and at the same time providing an attractive service to customers is a difficult challenge that every day is becoming more complex to address.
Some factors of the service/financial optimization equation are notably the global pandemic in addition to the difficulty in finding operational optimization scenarios, mainly in those terminals where almost all processes are already optimized and finally the pressure on the main game players to widen the differentiating gap with their competitors, especially in the technological area.
The objective of this article is to highlight some real applications of Deep Learning and Machine Learning (ML) in the operation and management of terminals and how with the help of AI they can find new sources of optimization to energize their own operation as well as the Colombian logistic chain.
To preface, this article will not go into details of algorithms, models, or programming languages to implement AI. The content is aimed at highlighting from a high-level view the following operational aspects that can benefit from AI and arouse curiosity in those who deal with the operation of a container terminal day by day.
Due to costs and business dynamics, importers want to have their cargo delivered as soon as possible. There are scenarios where AI can assist terminals in reducing the delivery time of import cargo as the following:
Let’s imagine that a terminal could have a deep learning model to predict the outcome of the customs clearance (physical inspection, documentary inspection, automatic clearance, etc) from the moment the cargo is announced. This could be a fantastic tool since terminals would know accurately and days in advance, how many inspections must be carried out and the number of containers that will be released without inspections. Having this information provides new tools for decision making in operational planning that each terminal will take advantage of its own experience and expertise.
The tool described above has as its final element a model known as a multi-class classifier. However, the complete solution landscape will articulate several models and data pipelines for example natural language processing (NLP) would be used to infer product codes from raw text cargo description. The good thing is that most of the terminal operators in Colombia have enough information in terms of parameters and volume to make possible this type of solution.
This is another scenario that could start from the moment a container is announced or discharged. In that scenario, a model could predict the delivery date (Just like Rappi and Amazon do) and notify the owner and other participants in the logistical process through one of the communication channels. Owners or customs brokers knowing the estimated delivery date can coordinate in advance with inland carriers on truck availability and speed up the process of getting the cargo to the final destination.
The model described in this scenario is known as a Regressor, it is a little more challenging than the previous one since the dataset must collect variables from the entire chain of operations that are executed from the moment the container is discharged until it leaves the terminal, steps that are often manual, such as manual data entry into information systems. This last point about manual interaction makes the construction of the model a bit tricky since the dataset must be able to collect factors such as the speed and experience of the human agents acting in the operation. Therefore, the delivery date may need to be re-inferred during some of the key moments of the container’s stay in the terminal.
Based on the foregoing, an additional benefit of having the cargo estimated delivery date is to be able to build a tool to detect hot spots in the import operation by highlighting that cargo whose delivery date deviates significantly from the estimates, this information can be displayed in a dashboard where users can drill down into individual cases and discovery hidden inefficiencies.
Once we have a model capable of predicting with high accuracy the cargo delivery date, a new tool can be created aimed at optimizing the business and operation of land transport. Going into details, the cargo delivery information of each terminal could be aggregated in a collaborative portal where inland transporters could browse and filter it by several categories (like the city of destination, weight, container size) and even bid on transport services. The main benefit for transporters is to be able to define intelligent routes and reducing idle time.
To conclude, applications of AI in terminals are even more extensive and include some areas such as predictive maintenance and fault prediction. Also, other applications are:
Convolutional networks trained to detect anomalies in container X-ray scan images.
Check-in wait time prediction.
Intelligent and automatic generation of slots to attend truck appointments.
If you want to know more about AI in ports, you can reach out to me at carlos.ortiz@k4s.co
Thanks to Lifeth Álvarez
After writing a series of articles on how to apply IA to recruiting, I decided it was time to take a break and change the topic of my articles (at least temporarily).
This article is the continuation of a series of posts covering applying AI and semantic search to tech recruitment. You can read them here:
Have you ever wanted a sourcing platform where you could start a recruitment workflow with this text:
“Startup is looking for a founder engineer with experience in Ethereum and smart contracts. Experience in frontEnd development with Angular is also desired.”
