Novosibirsk State Technical University (NSTU) is creating a platform for optimization, analysis and calculations of transport and warehouse logistics based on artificial intelligence and computer vision. The project will reduce the time required to perform basic logistics operations and optimize their number; as a result, it will reduce the cost and improve the quality of services provided.
According to the developers, transport logistics is one of the most resource—intensive industries with huge losses due to manual processes, lack of transparency, and planning errors. Existing optimization solutions are often fragmented, expensive, and do not use the full potential of modern technologies.
"We had the idea to create an intelligent solution. The task is to develop a single platform that automates key processes: vehicle accounting, loading/unloading control, cargo tracking, document management; analyzes large amounts of data in real time; optimizes key indicators such as routes, vehicle loading, fleet usage, fuel and repair costs, delivery times; increases transparency and security at all stages of the supply chain," said Egor Dorogin, a student at the Faculty of Applied Mathematics and Computer Science.
As part of the project, the development team creates its own artificial intelligence model, trains it to recognize vehicle numbers, body types, containers, pallets, and cargo. Video from cameras at warehouse gates, loading docks, in cabins, along routes is planned to be processed locally for initial analysis and compression, obtaining basic model parameters, after which AI is connected.
The uniqueness of the project lies in the combination of AI and computer vision (CV). Egor Dorogin provides several examples of how the platform works. An example with loading optimization: The CV recognizes that the truck has entered the warehouse, records the number and body type, then the AI checks with the order and determines what needs to be loaded. If the movers put the wrong pallets, the system immediately warns the dispatcher. An example with a route: The CV in the cab records that the driver is tired, the AI suggests the nearest point for rest and at the same time recalculates the route so that the delivery does not fail. And an example with cargo control: the CV scans the cargo at the departure of the car from the warehouse, compares it with the photo during loading, if there is damage, the AI automatically creates an act and notifies the customer.
The developers have identified several advantages of their project in comparison with their counterparts. These include full automation of data collection (no manual input is needed, since the CV records events itself), deep optimization (not just GPS tracking, but AI that rebuilds routes in real time), complexity (combining telematics, CV, planning and analytics in one system), as well as flexibility — you can connect only the necessary modules (for example, only number recognition, route optimization or vehicle composition, etc.).
The application is currently under development, the terms of reference, methods and algorithms are still being clarified and adjusted, and a prototype application is being designed and created. It is necessary to refine the CV accuracy, predictive analytics, conduct platform testing, and train AI. In the future, the team plans to integrate with various companies.