Vehicle Control Support via Centralized Learned Model Transmission
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle control systems face a performance gap between vehicles equipped with machine learning devices and those without, due to the high cost and computational demands of machine learning-based control systems, limiting the number of vehicles that can utilize learned models for efficient control.
Innovation Solution
A control support system that generates and transmits learned models using sensor information from multiple vehicles, allowing vehicles without machine learning devices to receive optimized control parameters, enabling equivalent control to that of vehicles with machine learning devices through a centralized server that performs machine learning and selects suitable learned models based on travel history information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning-based control systems are deployed in vehicles, then control precision and efficiency are improved, but device cost and computational requirements increase significantly
Solution Approach 1:
A server acts as an intermediary between data collection and model deployment. The server performs machine learning computations and generates learned models, which are then transmitted to vehicles. This mediator approach allows vehicles to benefit from sophisticated control algorithms without requiring complex on-board machine learning hardware.
Solution Approach 2:
Instead of deploying complex machine learning systems in each vehicle, the patent creates learned models (copies of the learned control patterns) that are transmitted to vehicles. These models contain the essential control knowledge without the computational overhead of the full machine learning system, enabling precise control with simpler device configurations.
2Measurement precision
If machine learning devices are installed in all vehicles, then control performance is improved, but manufacturing cost increases
Solution Approach 1:
The patent creates learned models that are transmitted to vehicles, serving as lightweight copies of the control knowledge. These models can be deployed across multiple vehicles without requiring expensive machine learning hardware in each unit, significantly reducing manufacturing costs while maintaining control performance.
Solution Approach 2:
A single server performs machine learning for multiple vehicles, providing a universal control solution. The learned models generated by the server can be applied across different vehicle instances, eliminating the need for each vehicle to have its own dedicated machine learning device, thereby reducing overall manufacturing costs.
3Measurement precision
If learned models are transmitted to vehicles, then control efficacy is improved in unequipped vehicles, but data transmission requirements increase
Solution Approach 1:
The patent extracts only the essential control parameters and learned model data needed for effective control, transmitting only this critical information to vehicles. This selective extraction approach ensures that control efficacy is improved in unequipped vehicles while minimizing the volume of data that needs to be transmitted.
Data Source
AI summary
A control support device for supporting control of a vehicle using a learned model obtained by machine learning, includes: a data acquisition unit acquiring sensor information, which is related to a state of an inside or an outside of a supplying vehicle that supplies parameters to be used for the machine learning; a learning unit generating a learned model by performing the machine learning using an input/output data set, which is the sensor information acquired by the data acquisition unit and is data including input parameters and an output parameter of the learned model; and a transmission unit Transmitting at least one of the generated learned model and an output parameter calculated by inputting sensor information of the vehicle, control of which is supported, to the generated learned model as an input parameter.