Distributed learning to learn context-specific driving patterns

This invention describes a system for self-driving cars to improve their navigation in specific locations or situations. Each vehicle generates an artificial intelligence model tailored to a particular driving context, such as a complex intersection or a specific road type. When a vehicle approaches such a location or another vehicle, it can share its learned driving model with other cars or roadside units, or receive models from them, to optimize driving for that specific environment. The claims narrow this sharing to situations where sufficient network bandwidth is available or when vehicles belong to the same fleet.