MIPT: Mathematical model of automated warehouse

2024: Introduction of a mathematical model of an automated warehouse

MIPT In created a mathematical model of an automated warehouse. The university announced this on June 26, 2024.

The algorithm proposed by the engineers of the Center for Applied Systems of Artificial Intelligence MIPT allows you to reduce the number of robots used by an average of a third, as well as guaranteed to improve the overall efficiency of warehouse processes by 10%. The development is focused on the needs of large business: the positive effect of its use in digital retail can reach several million rubles a month.

Warehouse automation provides many advantages: from optimizing the space used and increasing the overall speed of work to the obvious exclusion of the "human factor." At the same time, a person in a warehouse is still needed, and his job is not to carry loads, but to control robots, and sometimes to repair them.

Roman Gorbachev, head of the laboratory of wave processes and control systems at the MIPT Center for Applied Artificial Intelligence Systems, emphasizes that in order to effectively cope with professional tasks, a specialist will need not only knowledge in the field, logistics but also skills in working with algorithm systems computer. neuronets

If the robot breaks down and cannot go (for example, the bearing has worn out), then a person should intervene here. He has to take it out of there or push it out so it doesn't get in the way of other robots. Or I must inform all robots that this intersection has ceased to be passable. All cars will go around, packing food, moving shipping orders, accepting new goods, - said Roman Gorbachev.

According to Mikhail Zaripov, a leading programmer-developer of the laboratory of wave processes and control systems, interest in warehouse automation with the rapid growth of the logistics segment and the lack of specialists is quite expected - this allows you to optimize work processes and significantly save resources.

Based on the processing of the movement of agents (how they bypass each other, collide), the algorithm begins to draw highways. Highways on which you do not need to walk and the optimal ways of movement. Our algorithm analyzes the data and with the most non-standard warehouse configuration, with any number of agents, it will draw these corridors. It takes time to adjust the processes: it is necessary that robots repeatedly overcome the route. We guarantee a 10 percent increase in performance, - said Mikhail Zaripov, a leading programmer-developer of the laboratory of wave processes and control systems of the Center for Applied Systems of AI MIPT.

The algorithm proposed by MIPT engineers is able to mathematically calculate the optimization peak - the moment at which the best performance of a specific warehouse is achieved. This, in turn, allows not only to organize processes in an already operating warehouse with high efficiency, but also to adjust the infrastructure to specific parameters of the room where it is planned to place storage.

Our task is to find the optimal warehouse configuration. We came up with an algorithm that automatically goes through all the options and optimizes the processes for a specific location that we chose for the warehouse. So far, the owner only finds an empty room and estimates what and how he will fit in. (mount, robot paths, cabinets and tables). Automatic optimization and planning of the optimal warehouse. These are not neural networks, but machine learning algorithms. The closest thing to pick up is an ant algorithm (it controls collective actions in an ant colony). This natural mechanism was taken as the basis for the creation of our algorithm - the leading programmer-developer of the laboratory of wave processes and control systems of the Center for Applied Systems of AI MIPT Alexey Starostenko.

A mathematical algorithm for automating warehouse processes will become part of the technology framework - a set of IT tools that simplifies the development and support of complex robotic systems. In June 2026, MIPT employees are working on a new version of it. Also, the warehouse process optimization algorithm is applicable for any competitive logistics, and it can also be used to model transport infrastructure in the city and in industries. It can be a coal pit or a network of multi-lane highways.