Hardwired MLP Neural Layer Unit for Real-Time Control Calculation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in calculating precise mathematical models for complex systems like internal combustion engines within real-time requirements due to high computing demands, especially with data-based models requiring extensive resources.
Innovation Solution
A hardware-based model calculating unit for multilayer perceptron models with a hardwired processor core that calculates neural layers efficiently, using weighting matrices, offset values, and activation functions, allowing for real-time processing with reduced software load and lower resource requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If data-based models with a great number of supporting points are used to achieve sufficient modeling precision, then modeling precision is improved, but computing capacity requirements increase significantly
Solution Approach 1:
The patent segments the computational workload by dividing the great number of supporting points into multiple subsets, each processed by a separate neural network component or processing unit. This allows the system to handle large datasets through modular computation, reducing the computing capacity burden on any single unit while maintaining overall modeling precision.
Solution Approach 2:
The patent replaces traditional mechanical computation methods with neural network-based computational mechanisms. By using neural networks to process the supporting points, the system achieves efficient computation of complex relationships among multiple input variables, reducing the overall computing capacity requirements compared to conventional approaches.
2Measurement precision
If complex data-based models are calculated in real-time control unit applications, then modeling precision is improved, but real-time calculation capability deteriorates due to high computing demands
Solution Approach 1:
The patent performs preliminary computation by pre-processing and organizing the supporting points into neural network structures before real-time operation. The neural network is trained offline with the great number of supporting points, so that during real-time control applications, the pre-trained network can quickly infer results without requiring intensive real-time computation, thus achieving both high precision and real-time performance.
3Power
If a hardware-based model calculating unit is implemented to reduce computing capacity requirements, then computing capacity is reduced, but device complexity increases
Solution Approach 1:
The patent designs the neural network-based model calculating unit to be multi-functional, capable of processing various types of input variables and modeling different complex relationships through the same hardware architecture. This universality allows a single device to handle multiple tasks, reducing the need for multiple specialized components and thereby limiting the increase in device complexity.
Data Source
AI summary
A model calculating unit for calculating a neural layer of a multilayer perceptron model having a hardwired processor core developed in hardware for calculating a definitely specified computing algorithm in coupled functional blocks. The processor core is designed to calculate, as a function of one or multiple input variables of an input variable vector, of a weighting matrix having weighting factors and an offset value specified for each neuron, an output variable for each neuron for a neural layer of a multilayer perceptron model having a number of neurons, a sum of the values of the input variables weighted by the weighting factor, determined by the neuron and the input variable, and the offset value specified for the neuron being calculated for each neuron and the result being transformed using an activation function in order to obtain the output variable for the neuron.