Yandex robots start delivering faster using neural networks

Yandex has implemented a specialized transformer neural network in its delivery robots to plan routes and predict traffic conditions. Thanks to the new system, delivery speeds have increased by an average of 10 percent. This is reported by Ixbt.com .

The algorithm analyzes the behavior of pedestrians, cyclists, and vehicles. This allows robots to make fewer stops and choose more efficient movement trajectories. While robots previously relied on classical algorithms, the neural network now enables them to continue moving in complex situations while maintaining safety.

Company representatives state that the new system works in conjunction with previous algorithms. Each system proposes its own route, and the final decision is made based on the most optimal option. Additionally, robots have been equipped with special sound signals to warn of their approach.

The new technology is already being used in all delivery robots. Robots cover nearly 80 percent of their path under neural network control. Importantly, all calculations are performed directly on the robot's onboard computer, independent of an internet connection.