Online Learning Smart Steering for Wireless Mesh Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Wireless mesh networks face inefficiencies due to the 'sticky client' problem, where clients persistently connect to suboptimal access points, leading to less-than-optimum connection quality and bandwidth starvation in applications like video and gaming, as existing solutions lack adaptiveness and scalability to handle non-stationary network conditions.
Innovation Solution
A cloud-based, data-driven machine learning approach for smart client steering, utilizing a binary classification model trained on features like RSSI and cost metrics to predict successful steering actions, enabling real-time, computationally efficient, and adaptive decision-making for steering clients to better access points, addressing non-linear classification problems and scalability issues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Stability of the object's composition
If clients persistently connect to their initial access point, then connection stability is maintained, but network resource utilization deteriorates and bandwidth starvation occurs
Solution Approach 1:
The system implements a feedback mechanism where the machine learning model continuously learns from steering outcomes and link metrics to predict successful steering actions. The model receives feedback about whether previous steering attempts succeeded or failed, and uses this to improve future predictions, creating a closed-loop system that adapts to changing network conditions while maintaining connection stability.
Solution Approach 2:
The system changes the parameter of client connection behavior by using ML-predicted steering actions to dynamically redirect clients to more suitable access points. Instead of maintaining static connections, the system modifies connection parameters (target AP selection) based on predicted success probabilities and observed link metrics, thereby improving network resource utilization while preserving stability through intelligent selection.
2Productivity
If centralized steering management is implemented, then AP selection is optimized, but system complexity and computational overhead increase
Solution Approach 1:
The patent introduces a cloud-based machine learning service as an intermediary between the mesh network and clients. This intermediary handles the complex steering decisions by providing prediction APIs that network entities can query. The ML model runs independently in the cloud, receiving link metrics and returning steering predictions, thereby separating the optimization intelligence from the network infrastructure and reducing system complexity.
Solution Approach 2:
The system enables self-service by allowing network entities (gateways, APs, clients) to independently query the ML prediction service and make steering decisions based on returned predictions. Each entity uses the predicted success probabilities and link metrics to autonomously determine steering actions without requiring complex centralized coordination, thereby optimizing AP selection while minimizing system complexity.
3Ease of manufacture
If traditional steering methods are used, then implementation is simple, but adaptiveness to non-stationary network conditions deteriorates
Solution Approach 1:
The system implements dynamics by using a machine learning model that continuously adapts to changing network conditions. The ML model learns from observed link metrics and steering outcomes, dynamically adjusting its predictions to reflect current network state. This allows the system to adapt to non-stationary conditions (changing traffic patterns, client mobility, AP performance) while maintaining implementation simplicity through the use of established ML techniques and cloud-based deployment.
4Adaptability or versatility
If online learning is implemented, then adaptiveness to real-time conditions is improved, but computational requirements increase
Solution Approach 1:
The patent uses a cloud-based ML service as an intermediary to handle computationally intensive online learning tasks. The heavy computational workload of training and updating the ML model occurs in the cloud infrastructure, which has abundant computational resources. Network entities only need to perform lightweight operations: collecting link metrics, querying the prediction API, and acting on returned predictions. This distribution of computational requirements enables real-time adaptiveness without burdening the mesh network devices.
Data Source
AI summary
An online learning-based smart steering system and method comprises an online kernel perceptron classifier. The smart steering approach/method performs learning sequentially at the cloud from the entire data of multiple mesh networks, and, operates at APs for steering; both of which are executed in real-time. The system and method uses network features such as current RSSI (received signal strength indicator), current cost, target RSSI and target cost for every steering action resulting in either success of failure. The system and method uses a classifier based on this data for obtaining a model in the feature space which predicts whether a steering action can succeed or not before it is issued.