Neural Network Bin Selection for Time-Triggered Ethernet Scheduling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current algorithms for Time-Triggered Ethernet (TTE) network scheduling fail to converge on a valid global schedule for proper traffic throughput due to low-rate, low-latency data sets that violate assumptions about rate and latency requirements, leading to difficulties in scheduling Virtual Links (VLs).
Innovation Solution
Implementing a Multi-Agent Deep Deterministic Policy Gradient (MADDPG) approach using neural networks for optimizing bin selection in network scheduling, which trains through multiple iterations and data sets to adjust the time slot selection algorithm policy, treating each bin as an independent agent and calculating reward functions based on available bins, VLs, and scheduling success.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If current algorithms are used for TTE network scheduling, then the scheduling process is simple, but the algorithms fail to converge on a valid global schedule for low-rate, low-latency data sets
Solution Approach 1:
The patent replaces traditional mechanical scheduling algorithms with a neural network-based machine learning system. The neural network learns optimal bin selection strategies through training on historical scheduling data, substituting deterministic algorithmic rules with probabilistic, data-driven decision-making that adapts to low-rate, low-latency traffic patterns
Solution Approach 2:
The patent transforms the scheduling problem by changing the parameter space from direct time slot assignment to bin selection based on learned features. The neural network processes multiple input parameters (traffic rate, latency requirements, network topology) and outputs optimized bin selections that traditional algorithms cannot determine, effectively changing how scheduling decisions are made
2Productivity
If traditional bin selection methods are used, then the scheduling algorithm is easy to implement, but bin packing efficiency is insufficient for complex traffic patterns
Solution Approach 1:
The patent applies preliminary action by training the neural network offline on extensive scheduling data before deployment. The network learns optimal bin selection patterns in advance through multiple training iterations, so that during actual network operation, bin selection decisions can be made quickly based on pre-learned knowledge rather than complex real-time calculations
Solution Approach 2:
The neural network serves as an intermediary between the scheduling algorithm and the bin selection process. Instead of the algorithm directly selecting bins based on simple rules, the neural network acts as a intelligent mediator that processes traffic characteristics and network state, then recommends optimal bin selections that improve packing efficiency
3Reliability
If simple scheduling algorithms are used, then the system is easy to operate, but low-rate, low-latency traffic cannot be effectively scheduled
Solution Approach 1:
The neural network implements self-service by automatically learning and adapting to different traffic patterns without requiring manual reconfiguration. The system trains on historical data and autonomously improves its bin selection capabilities, eliminating the need for operators to manually adjust scheduling parameters for different traffic types like low-rate, low-latency flows
Data Source
AI summary
Methods, systems and apparatuses to enable an optimum bin selection by implementing a neural network with a network scheduling and configuration tool (NST), the method includes: configuring an agent with a critic function from neural networks wherein the agent neural network represents each bin of the collection of bins in the network that performs an action, and a critic function evaluates a criteria of success for performing the action; processing, by a scheduling algorithm, the VLs by the NST; determining one or more reward functions using global quality measurements based on criteria comprising: a lack of available bins, a lack of available VLs, and successfully scheduling operations of a VL into a bin; and training the network based on a normalized state model of the scheduled network by using input data sets to arrive at an optimum bin selection.