Vehicle Auto Reply Architecture for Fast and Detailed Responses
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle occupant interaction systems fail to provide appropriate responses to utterances, lacking detailed and timely replies that satisfy the occupant's preferences.
Innovation Solution
An auto reply system utilizing a first generation model on the vehicle and a larger second generation model on a server, generating and combining first and second reply information to provide detailed and timely responses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large-scale generation model is used to generate detailed reply information, then the response accuracy and detail are improved, but the processing time and system complexity increase
Solution Approach 1:
The system divides the generation model into two segments: a first generation model deployed in the vehicle for quick processing and a second generation model deployed on the server for detailed processing. This segmentation allows the system to balance between response speed and accuracy by handling different types of queries with appropriately sized models.
Solution Approach 2:
The system transitions from a single-model architecture to a distributed two-model architecture across different spatial locations (vehicle and server). This dimensional change in system architecture enables simultaneous operation of models with different scales, resolving the contradiction between speed and accuracy.
2Measurement precision
If a large-scale generation model is deployed in the vehicle, then the response accuracy is improved, but the device complexity and computational requirements increase
Solution Approach 1:
The system segments the computational workload by deploying a lightweight first generation model in the vehicle and a larger second generation model on the server. This segmentation reduces the complexity burden on the vehicle's onboard systems while maintaining access to advanced processing capabilities through the server.
Solution Approach 2:
The server acts as an intermediary that provides access to the large-scale second generation model without requiring the vehicle's onboard system to directly host it. This intermediary approach allows the vehicle to benefit from high-accuracy processing while keeping its own system complexity manageable.
3Loss of information
If the system waits for the second reply information from the server, then the response detail is improved, but the waiting time increases
Solution Approach 1:
The system performs preliminary action by generating first reply information using the first generation model before receiving the second reply information from the server. This preliminary response is provided to the occupant immediately, and the more detailed second reply information is integrated afterward, ensuring both timely and comprehensive information delivery.
Solution Approach 2:
The system maintains continuity of useful action by continuously providing reply information to the occupant - first with the immediate first reply information, then supplementing it with the more detailed second reply information. This continuous information delivery ensures the occupant receives useful responses without unnecessary waiting.
Data Source
AI summary
An auto reply device included in an auto reply system and mounted on a vehicle generates first reply information by inputting input information including voice information representing an utterance of an occupant of the vehicle into a first generation model that is pre-trained to generate the first reply information, generates inquiry information representing the utterance, based on the input information, and replies to the occupant, based on at least one of the first reply information and second reply information generated based on the inquiry information in a server provided outside the vehicle. The server generates the second reply information by inputting the inquiry information received from the auto reply device into a second generation model that is pre-trained to generate the second reply information and larger than the first generation model.