Furcifer Middleware for Adaptive Local-Edge-Split Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing computing configurations for real-time applications, such as object detection, struggle to adapt dynamically to highly dynamic environments with unreliable connectivity and varying system loads, leading to inefficiencies in energy consumption, performance, and latency.
Innovation Solution
Furcifer, a middleware framework that transparently monitors system resources and predicts the feasibility of edge, local, and split computing configurations, seamlessly switching between them to optimize energy consumption and performance using container-based services and low-complexity predictors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If neural algorithms are deployed at mobile devices, then performance is improved, but energy consumption increases
Solution Approach 1:
The patent implements dynamic configuration switching between local, edge, and split computing modes based on real-time system state assessment. The Furcifer middleware continuously monitors network conditions, device resources, and task characteristics to adaptively select the optimal computing configuration, enabling the system to transition from static to dynamic operation and resolve the performance-energy tradeoff.
Solution Approach 2:
The system changes operational parameters by adjusting the degree of computation offloading and model splitting based on environmental conditions. By varying parameters such as network bandwidth availability, device computational capacity, and task priority, the system optimizes the balance between performance and energy consumption across different operating scenarios.
2Use of energy by moving object
If neural models are offloaded to edge servers, then energy consumption is reduced, but latency increases due to data transfer
Solution Approach 1:
The patent applies split computing by dividing the neural network model into segments, with certain layers executed locally on the mobile device and other layers executed on the edge server. This segmentation reduces the amount of data that needs to be transmitted over the network, thereby reducing latency while still achieving energy savings through partial offloading.
Solution Approach 2:
The Furcifer middleware acts as an intermediary that intelligently manages the distribution of computational tasks between the mobile device and edge server. It assesses system state and dynamically determines the optimal split point for model execution, mediating the tradeoff between local processing (lower latency) and remote processing (lower energy consumption).
3Device complexity
If computing configuration is statically set, then system complexity is reduced, but adaptability to dynamic environments deteriorates
Solution Approach 1:
The Furcifer middleware implements self-service by autonomously monitoring system state, assessing feasibility of different computing configurations, and switching between local, edge, and split computing modes without external intervention. The system automatically adapts to changing network conditions and device states, providing context-aware configuration selection that enhances adaptability while maintaining manageable complexity through automated decision-making.
4Measurement precision
If high-performance neural models are used, then accuracy is improved, but computational resource requirements increase
Solution Approach 1:
The patent segments high-performance neural network models into multiple layers or components, allowing the system to selectively execute computationally intensive segments on edge servers with sufficient resources while performing lighter segments locally. This segmentation enables the use of accurate high-performance models without requiring the entire model to run on resource-constrained mobile devices.
Solution Approach 2:
The system creates a universal computing framework that can execute different model configurations across multiple platforms (mobile devices and edge servers). By designing the architecture to support both local and remote execution of model segments, the system achieves multi-functionality that allows high-accuracy models to be utilized regardless of the specific device's computational resources.
Data Source
AI summary
The technology disclosed provides Furcifer: a framework capable of dynamically adapting the cloud continuum computing configuration in response to the perceived state of the system. Our container-based approach incorporates low-complexity predictors that generalize well across operating environments. In addition, we develop a highly optimized split Deep Neural Network model, which achieves in-model supervised compression and enhances task offloading. Experimental results for object detection across diverse conditions, environments, and wireless technologies, show Furcifer's remarkable outcomes, including a 2× energy reduction, 30% higher mean Average Precision score than pure local computing, and a notable three-fold increase in frame per second rate compared to static offloading.