Autonomous Vehicle Scenario Modules for Shared Control Decisions

The patent segments the autonomous vehicle control system into distinct scenario-specific modules (e.g., pedestrian crossing scenarios, intersection scenarios, highway merging scenarios). Each module contains specialized control strategies and evaluation models tailored to specific operational contexts, allowing the system to handle complexity through modular organization rather than monolithic design.

The patent introduces an external shared scenario-specific operational control management system that acts as an intermediary. This external system receives operational data from multiple autonomous vehicles, processes it centrally, and returns refined control strategies. This mediator handles the computational burden of aggregating and analyzing data from multiple sources, reducing the complexity burden on individual vehicle systems.

The patent implements pre-processing and filtering of operational experience data before it is shared across the network. Each vehicle pre-processes its own operational data to extract only the most relevant and valuable insights, reducing the volume of data that needs to be transmitted and processed in real-time. This preliminary action ensures that only high-quality, decision-critical information is shared, maintaining accuracy while reducing communication overhead.

The patent uses simplified representations and models of operational scenarios rather than transmitting complete raw sensor data. Each vehicle creates compressed, scenario-specific copies of its operational experience that capture essential patterns and outcomes. These copied representations are much smaller and faster to process while retaining the decision-making value of the original full-resolution data.

The patent implements dynamic selection and instantiation of scenario-specific control evaluation models based on the current operational context. Rather than running all possible scenario models continuously, the system dynamically activates only the models relevant to the current situation (e.g., activating pedestrian crossing models only when approaching a crosswalk). This dynamic approach maintains high control precision for relevant scenarios while dramatically reducing overall computational energy consumption.

The patent applies specialized, high-precision control evaluation models only to specific local scenarios where they are most needed, rather than applying uniform high-computation models to all situations. Each scenario-specific model is optimized for its particular context (e.g., intersection models focus on right-of-way logic, while highway models focus on speed and spacing), providing locally optimal control precision without the energy cost of running all models everywhere.