RAN Application Conflict Coordination Using Reinforcement Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radio access network (RAN) technologies face challenges in managing conflicts among multiple applications due to dynamic and unpredictable interactions, leading to sub-optimal performance, as current approaches rely on static policies that fail to scale with the addition of new applications.
Innovation Solution
Implementing a reinforcement learning-based algorithm to identify and resolve conflicts among RAN applications by selecting optimal execution modifications, such as skipping, postponing, or reordering requests, using an RL-based agent trained to maximize performance indicators and minimize conflicts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If static policies are used to manage RAN applications, then the system is simple to implement, but the system cannot scale with new applications and performs sub-optimally
Solution Approach 1:
The patent applies dynamics by replacing static policies with a reinforcement learning-based dynamic policy. The RL agent continuously learns from network state observations and adjusts execution requests in real-time, enabling the system to adapt to new applications and changing conditions without manual reconfiguration. This dynamic approach allows the system to scale with new applications while maintaining optimal performance.
Solution Approach 2:
The system implements self-service through the reinforcement learning agent that autonomously manages application conflicts and executes requests. The RL agent independently learns from network data, makes decisions about request prioritization and execution, and continuously improves its policy without human intervention. This self-managing capability enables scalability while reducing operational complexity.
2Adaptability or versatility
If multiple RAN applications execute simultaneously, then network functionality is enhanced, but conflicts arise leading to sub-optimal performance
Solution Approach 1:
The patent applies preliminary action by having the reinforcement learning agent analyze and prioritize execution requests before they are executed. The agent evaluates potential conflicts between applications in advance and pre-determines optimal execution orders, ensuring that conflicting requests are resolved before execution begins. This proactive approach maintains high functionality while preventing performance-degrading conflicts.
Solution Approach 2:
The system implements feedback through the reinforcement learning loop that continuously monitors network performance and uses this information to update the agent's policy. The agent receives feedback from network state observations and execution outcomes, allowing it to learn from past decisions and improve future request management. This feedback mechanism ensures optimal performance while enabling multiple applications to execute simultaneously.
Data Source
AI summary
A processing system may receive a set of execution requests from a plurality of radio access network applications within a defined period of time, where each execution request of the set of execution requests is to perform at least one execution for configuring at least one aspect of a radio access network. The processing system may next identify at least one conflict among at least two of the plurality of radio access network applications. The processing system may then select, in response to identifying the at least one conflict, at least one modification to the set of execution requests in accordance with a reinforcement learning-based algorithm, and may perform at least a first execution associated with the set of execution requests according to the at least one modification to configure at least a first aspect of the radio access network.