Yandex Upgrades Delivery Robots With Transformer-Based Neural Network
The new system continuously adjusts rover routes without stopping the vehicles.
Yandex has rolled out a major software update for its autonomous delivery robots, replacing traditional route-planning algorithms with a transformer-based neural network. The company says the system can guide delivery rovers through crowded urban environments filled with pedestrians and vehicle traffic.
Previously, rover navigation relied on conventional algorithmic planners. When robots entered high-traffic areas, the system often slowed them down or stopped them entirely until obstacles cleared. The new neural-network planner keeps the robots moving continuously, helping increase delivery speeds by roughly 10%, according to the company.
The AI planner was trained using simulated driving sessions with a combined virtual runtime equivalent to 4,000 years. To generate routes, the system processes data collected from the robot’s onboard sensors. All computations are performed directly on the device itself, allowing the robots to continue operating even under mobile internet restrictions.
