Task Assignment via Event Density Spectral Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for automatic synthesis of distributed embedded systems often result in suboptimal solutions due to insufficient consideration of time limits, leading to potential product recalls and increased costs, as they fail to efficiently map tasks to hardware architecture while meeting real-time criteria and minimizing hardware requirements.
Innovation Solution
A computer-implemented method that assigns tasks to processing units based on data interdependencies and event densities, using spectral analysis and task partitioning to ensure predefined time limits are met, while optimizing costs through energy consumption, power, and chip area considerations, employing mathematical real-time calculus and scheduling methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If tasks are assigned to processing units using conventional optimization methods, then hardware architecture can be minimized, but predefined time limits of tasks cannot be reliably met
Solution Approach 1:
The patent applies preliminary action by performing spectral analysis and event density calculations on task graphs before final task assignment. The method pre-calculates event densities, identifies critical paths, and determines task partitioning strategies in advance, ensuring that time limit constraints are built into the assignment process from the beginning rather than being added as post-processing checks.
Solution Approach 2:
The patent segments the task assignment process into distinct phases: spectral analysis of event densities, task graph partitioning, critical path identification, and constrained optimization. This segmentation allows each aspect (hardware minimization and time limit compliance) to be optimized independently and then integrated, resolving the contradiction between minimizing hardware and meeting time constraints.
2Device complexity
If tasks are assigned without considering event densities and data interdependencies, then assignment process is simplified, but hardware oversizing is necessary to meet time limits
Solution Approach 1:
The patent changes the parameters of task assignment by introducing event density metrics and spectral analysis. Instead of using simple heuristic assignment, the method transforms the assignment problem into a spectral domain analysis, calculating event densities and using these transformed parameters to guide optimal task-to-processing-unit mapping, thereby avoiding hardware oversizing.
Solution Approach 2:
The patent replaces conventional mechanical/optical optimization methods with mathematical spectral analysis. Instead of using iterative trial-and-error assignment methods, the invention substitutes a mathematical transformation approach that analyzes event densities in the spectral domain, providing a more efficient and accurate assignment process that avoids both oversimplification and excessive complexity.
3Productivity
If first-step block distribution to control devices is performed without detailed system architecture knowledge, then initial assignment is快速 completed, but suboptimal solutions result requiring complex optimisation measures
Solution Approach 1:
The patent applies preliminary action by performing comprehensive spectral analysis and event density calculations during the initial task assignment phase, rather than relying on simple block distribution followed by post-hoc optimization. This preliminary rigorous analysis ensures that the first assignment is near-optimal, eliminating the need for complex subsequent optimization measures while maintaining high assignment speed through efficient mathematical algorithms.
Data Source
AI summary
The present invention relates to a computer-implemented method for an automatic synthesis of distributed embedded systems, wherein the tasks to be processed by the system are mapped to a hardware structure having a plurality of processing units such that predefined time limits of the tasks are met, comprising the steps of (a) assigning the tasks to the plurality of processing steps, with the following substeps: (aa) assigning a task to a processing unit; (bb) determining the outgoing event densities; (cc) comparing the output density towards the next task with a predefined threshold and assigning the next task to the same processing unit if the event density is below the threshold or assigning the next task to any other processing unit if the event density is smaller than the threshold; (dd) repeating steps (aa) to (cc) until all tasks are assigned to the processing units; (b) checking whether the costs of the given task assignment to the processing units satisfy a predefined solution criterion; (c) repeating steps (a) to (b) with a new task assignment to the processing units until the task assignment fulfills the predefined solution criteria; (d) assigning the tasks to the processes of the operational systems of the processing units assigned to the tasks; (e) checking whether the given task assignment to the processes of the operational systems of the processing units satisfies the predefined time criteria of the tasks; (f) calculating the costs associated with the given task assignment to the processes of the operational systems of the processing units if the predefined time criteria of the tasks are satisfied; (g) repeating steps (a) to (c) with a new task assignment to the processing units or repeating steps (d) to (f) with a new task assignment to the processes of the operational systems of the assigned processing units until the costs of the current solution satisfy a predefined solution criterion.