Intermittent Computing Platform With Energy-Aware State Saving
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for processing data in the intelligence revolution, such as autonomous vehicles and real-time traffic management, are limited by the need for continuous power and energy supply, which is not compatible with intermittent energy sources, leading to longer execution times and incompatibility with energy availability.
Innovation Solution
A platform for intermittent computing that includes a non-volatile memory, microcontroller for power management, and a System-on-Chip (SoC) with reprogrammable logic, allowing applications to be executed intermittently based on available energy levels, with energy harvesting from various sources and automatic saving and restoring of application context.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If traditional processing approaches with continuous power supply are used, then computing power and processing speed are improved, but energy independence and adaptability to intermittent energy sources deteriorate
Solution Approach 1:
The system dynamically adapts its operation mode based on available energy levels. The processing unit switches between continuous operation (when energy is abundant) and intermittent operation with periodic saving (when energy is scarce), making the computing power adaptable to varying energy availability rather than requiring continuous supply.
Solution Approach 2:
The system changes operational parameters based on energy conditions. When energy becomes intermittent, the system modifies its behavior by periodically saving computation results to non-volatile memory and resuming later, effectively changing from a continuous processing parameter regime to an intermittent one that maintains progress despite power fluctuations.
2Manufacturing precision
If complex processing operations are executed continuously, then processing completeness is improved, but execution time exceeds available energy duration
Solution Approach 1:
The system performs preliminary saving of computation results at periodic intervals during processing. By periodically storing intermediate results in non-volatile memory before energy depletion occurs, the system ensures that processing progress is preserved and can be resumed, guaranteeing eventual completion even when total execution time exceeds continuous energy availability.
Solution Approach 2:
The system maintains continuity of useful computation through periodic resumption. Instead of losing all progress when energy becomes unavailable, the system saves state information and resumes computation from the last saved point, ensuring that the useful computational action continues accumulatively over time rather than being interrupted with total loss.
3Power
If data is processed at remote cloud data centers, then centralized processing capability is improved, but reaction time for rapid decision-making deteriorates
Solution Approach 1:
The system segments the centralized cloud processing model by implementing local edge computing capability. Instead of routing all data to remote data centers, the processing unit performs computations locally at the edge device, dividing the processing function between local intermittent computing and potential cloud synchronization, thereby reducing transmission delays and improving reaction time for time-critical decisions.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The disclosure notably relates to a platform (100) for intermittent computing powered by at least one energy source. The platform comprises a non-volatile memory (118). The platform comprises a microcontroller (120) configured for power management of the at least one energy source and for determining a level of available energy. The platform comprises a System-on-Chip (SoC) component comprising a reprogrammable logic (118). The reprogrammable logic comprises an application layer (202) comprising operational blocks implementing one or more applications. The reprogrammable logic comprises an intermittent computing layer (204) configured for executing intermittently the one or more applications of the application layer depending on the determined level of available energy by regularly saving and restoring information from the application layer on the non-volatile memory. Such a platform presents an improved solution of energy-independent and sustainable by design system capable of performing complex processing.