Russian Researchers Are Using AI to Cut Fuel Consumption in Heavy Trucks—Without Modifying the Engine
A new software system combines physics-based modeling and machine learning to reduce fuel use and emissions in long-haul tractors, relying only on vehicle data.
Scientists at Moscow Polytechnic University are developing a software-and-algorithmic system designed to reduce fuel consumption and carbon monoxide emissions in semi-trailer trucks. The solution does not require any changes to the vehicle’s physical design, TASS reported, citing the university’s press service.
The project is based on a hybrid approach that combines classical physics models with artificial intelligence. According to Artem Dubrovsky, the study’s lead author and a lecturer at the university’s Advanced Engineering School of Technological Leadership, the system first calculates vehicle motion and fuel consumption using physical parameters such as aerodynamics, rolling resistance, road gradients, and route profiles.
The model is then automatically adapted to a specific truck using data from real-world experiments.
Vehicle condition data is taken directly from the truck’s onboard network—the CAN bus—which connects sensors and control units. This allows the system to be quickly calibrated for a specific vehicle. Machine learning methods then refine the predictions based on real operating conditions, including speed, slope, road curvature, surface characteristics, and driving style.
Using this combined analysis, the system generates an optimal speed profile and provides the driver with recommendations. Crucially, the software does not just issue advice—it explains the reasoning behind each recommendation.
Dubrovsky emphasized that the system does not rely on AI alone. The physics-based model produces the initial calculation, which is then carefully adjusted by a machine-learning model trained on experimental data. Once trained, the system can be applied to new routes for the same vehicle. If conditions such as vehicle weight change, recalibration takes only a few minutes.
This makes the technology useful not only for fuel savings, but also for driver training and route analytics.
The software complex will be tested under laboratory conditions using real vehicle data. Earlier reports also noted that Russian scientists have found a way to “soften” the behavior of high-energy fuels—a development that could make future reactors safer and more efficient if confirmed in real-world conditions.
