Wi-Fi Contention Window Selection for Predictable Channel Access
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing Wi-Fi technologies in license-exempt bands face unpredictable channel access and high latency due to uncontrolled interference, leading to inefficiencies in channel usage and challenges in supporting applications with demanding quality of service (QoS) requirements.
Innovation Solution
Implementing a centralized channel access mechanism with AP coordination and deep reinforcement learning to dynamically select contention window sizes, using a Markov Decision Process and multi-agent reinforcement learning to optimize channel access for multiple access points (APs).
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If distributed channel access (DCF) is used, then device simplicity and ease of operation are improved, but channel access predictability and QoS support deteriorate under high load
Solution Approach 1:
A central controller is introduced as an intermediary to coordinate channel access for multiple APs. The controller receives channel access requests from APs, manages contention window sizes, and allocates transmission opportunities, thereby centralizing control to improve predictability while maintaining distributed architecture benefits
Solution Approach 2:
The system dynamically adjusts contention window (CW) sizes based on channel conditions and traffic load. By changing the CW parameter adaptively through reinforcement learning, the system optimizes channel access probability and reduces collisions, improving both predictability and throughput under varying load conditions
2Reliability
If centralized channel access with AP coordination is implemented, then channel access predictability and QoS support are improved, but system complexity increases
Solution Approach 1:
The system employs reinforcement learning algorithms that enable automatic, adaptive decision-making for CW size selection and channel access timing. The controller learns optimal policies through interaction with the environment, eliminating the need for manual configuration or complex centralized scheduling algorithms, thereby reducing operational complexity while maintaining high predictability
3Productivity
If dynamic CW selection using reinforcement learning is used, then throughput and channel utilization are improved, but computational overhead and processing requirements increase
Solution Approach 1:
The reinforcement learning implementation focuses on learning and optimizing only the critical CW size parameter rather than all channel access parameters simultaneously. This partial action approach concentrates computational resources on the most impactful decision, achieving significant throughput improvements with reduced processing overhead compared to full-parameter optimization
Data Source
AI summary
A method performed by a controller is provided. The method includes receiving first input information from a first network node. The first input information includes any one or more of: a first group of one or more values for one or more parameters associated with a first contention window, CW, value, a second CW value, and a second group of one or more values for the one or more parameters associated with the second CW value. The method further includes obtaining first policy information indicating a first policy which is determined based at least on the first input information and transmitting towards the first network node the first policy information. The first policy is for determining a third CW value for the first network node.