Edge AI Model Switching via Cloud Intermediary
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing information processing apparatuses struggle to dynamically switch between different artificial intelligence models based on varying inference processes and desired outcomes, especially in edge-side devices where conditions and requirements change over time.
Innovation Solution
An information processing apparatus that includes an information acquiring section to gather data from sensors, a model selecting section to choose the appropriate artificial intelligence model based on user-defined purposes and sensor information, and a transmission processing section to manage the deployment of selected models to edge-side devices, ensuring seamless updates and model switching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If manual updating of artificial intelligence models is performed, then model version control is maintained, but the ability to dynamically switch between different model types and versions is limited
Solution Approach 1:
A cloud-based AI model management system is introduced as an intermediary between the edge device and multiple AI model versions. The management system stores various AI models, manages version information, and communicates with the edge device to enable automatic model switching without increasing local device complexity. The edge device simply queries available models and receives updates from the cloud intermediary.
Solution Approach 2:
The system transitions from local-only model storage to a cloud-based dimensional expansion. Instead of managing models locally on the edge device, the system adds a cloud dimension where models are stored, versioned, and managed. This allows unlimited model versions to be maintained without increasing edge device complexity, as the cloud infrastructure absorbs the complexity of model management.
2Adaptability or versatility
If multiple AI models of different types are deployed, then inference process adaptability is improved, but the ease of operation and model selection becomes more difficult
Solution Approach 1:
The system implements feedback loops where the edge device sends usage information and inference results back to the cloud management system. Based on this feedback, the management system automatically determines which models to push to the edge device, eliminating manual selection. The feedback mechanism enables automatic adaptation to changing inference needs without user intervention.
Solution Approach 2:
The system enables self-service operation where the edge device automatically queries available AI models from the cloud, receives model updates without manual intervention, and switches between models based on predetermined conditions. The management system autonomously handles model versioning, compatibility checking, and deployment, freeing the user from manual model management tasks.
3Loss of time
If real-time model switching is enabled, then response time to changing requirements is reduced, but the reliability of the inference process may be compromised during transitions
Solution Approach 1:
The system performs preliminary actions by maintaining multiple AI model versions in the cloud management system ready for deployment. When a model update is needed, the new model is prepared and validated in advance on the cloud side before being pushed to the edge device. This preliminary preparation allows rapid deployment without compromising reliability, as the new model is already verified and ready for immediate use.
Solution Approach 2:
The system implements beforehand cushioning by maintaining a library of pre-validated AI models in the cloud, including previous versions that serve as fallback options. When switching models, the system has backup versions ready to prevent inference failures. This cushioning approach ensures that if a model transition goes wrong, the system can fall back to a previous stable version, maintaining reliability during rapid updates.
Data Source
AI summary
An information processing apparatus according to the present technology includes an information acquiring section that acquires information from a sensor of an edge-side information processing apparatus that performs an inference process using an artificial intelligence model, a model selecting section that selects an artificial intelligence model to be deployed on the edge-side information processing apparatus, on the basis of purpose information set by a user and the information acquired from the sensor, and a transmission processing section that performs transmission control for transmitting the selected artificial intelligence model to the edge-side information processing apparatus, on the basis of a confirmation result of purchase information regarding the selected artificial intelligence model in an account associated with the user.