Researchers at Chalmers University in Sweden have introduced a new optimization method for managing autonomous electric taxi fleets. The project, conducted in collaboration with Volvo Cars, utilizes a combination of machine learning and classical mathematical modeling to improve urban traffic efficiency.
The system employs Bayesian Neural Networks to forecast passenger demand and charging station availability. These models were trained on historical taxi data from San Francisco and Chicago. A key feature of this approach is the ability to estimate prediction uncertainty, which helps the system maintain stability during unexpected events like sudden weather changes.
The framework separates data analysis from operational control. While the AI provides the foundation, a mathematical model with fixed rules manages actual driving and charging schedules. To handle the complexity of large urban environments, the algorithm divides tasks into smaller segments, allowing for decision-making in fractions of a second.
The research team also analyzed three different electric vehicle configurations with varying battery sizes and efficiency levels. The study highlights that balancing battery capacity, charging logistics, and demand forecasting is essential for reliable fleet operation. According to Stanford University, autonomous electric taxis could reduce costs per kilometer by up to 60 percent compared to traditional taxis.
