Language Model Driving Plans for Local Traffic Rule Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional autonomous vehicles are limited to operating in geo-fenced areas due to their inability to adapt to varying traffic rules across different geographic locations, leading to potential safety concerns and reduced utility.
Innovation Solution
A computer-implemented method that uses a language model driving assistant to receive text descriptions of scenes and plans for vehicle driving, extract relevant traffic rules, generate prompts for driving instructions, and process these prompts to generate updated driving plans that adhere to local traffic rules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a machine learning model is trained to adhere to a specific set of traffic rules, then the model can control the vehicle according to those rules, but the vehicle cannot operate in locations with different traffic rules
Solution Approach 1:
The system dynamically adapts the machine learning model's behavior by injecting different traffic rule parameters based on the vehicle's current geographic location. The controller receives location information and modifies the model's input prompts accordingly, allowing the same model to adhere to different traffic rules in different regions without retraining.
Solution Approach 2:
The system changes the parameters fed to the machine learning model based on geographic location. By modifying the traffic rule parameters in the prompt input according to the vehicle's location, the model generates appropriate driving decisions for different jurisdictions while maintaining the same underlying model structure.
2Stability of the object's composition
If a machine learning model is trained with a specific set of traffic rules, then the model provides consistent control decisions, but erroneous control occurs when the vehicle enters areas with different traffic rules
Solution Approach 1:
The system incorporates location-based feedback into the machine learning model's decision-making process. The controller continuously monitors the vehicle's geographic location and adjusts the traffic rule parameters in real-time, ensuring the model receives accurate contextual information for safe decision-making in any location.
Solution Approach 2:
The system performs preliminary adaptation by pre-loading location-specific traffic rule parameters into the model's prompt before the vehicle enters a new geographic area. This proactive adjustment prevents erroneous control decisions by ensuring the model is already configured with the correct traffic rules for the upcoming location.
3Reliability
If the vehicle is limited to geo-fenced areas, then safety is maintained within those areas, but the utility of the autonomous vehicle is reduced
Solution Approach 1:
The system achieves universal applicability by enabling a single machine learning model to operate safely across multiple geographic locations with different traffic rules. The location-adaptive prompt injection mechanism allows the model to function correctly in diverse jurisdictions, eliminating the need for geo-fencing limitations.
Data Source
AI summary
One embodiment of a method for controlling a vehicle includes receiving first text that includes a description of a scene and a first plan for driving a vehicle, extracting at least one portion of a set of traffic rules based on the description of the scene and the first plan, generating a first prompt that requests driving instructions and includes the description of the scene, the first plan, and the at least one portion of the set of traffic rules, processing the first prompt via a first trained language model to generate a second plan for driving the vehicle, and generating driving instructions based on the second plan.