Coexistence Manager (CxM)

The CxM's primary function is to resolve contention when multiple General Authorized Access (GAA) users request overlapping spectrum resources. It ingests interference reports and applies a coordination algorithm to generate a conflict-free channel assignment plan. The process involves:

• Analyzing pairwise interference potential between all requesting CBSDs
• Applying Aggregate Interference Margin constraints to protect incumbents
• Iteratively adjusting power levels and frequency assignments until a stable solution is reached

A CxM must coordinate across different network operators and service providers within the same geographic area. It acts as a neutral arbiter, ensuring fair access without favoring any single operator. Key mechanisms include:

• Proportional Fairness Scheduling to balance total throughput with individual user data rates
• Managing inter-operator interference at administrative boundaries
• Enforcing Spectrum Etiquette rules when operators cannot directly negotiate

The CxM employs computationally intensive algorithms to solve the NP-hard channel assignment problem. Common approaches include:

• Graph coloring where nodes represent CBSDs and edges represent harmful interference
• Distributed Constraint Optimization (DCOP) for multi-agent negotiation
• Heuristic search techniques that converge on a near-optimal solution within strict latency budgets, typically under 30 seconds for a re-prioritization event

When a Dynamic Protection Area (DPA) is activated by a federal incumbent sensor, the CxM must immediately recalculate all assignments. This involves:

• Identifying all CBSDs within the DPA's neighborhood that contribute to aggregate interference
• Forcing suspension or power reduction for affected devices within 60 seconds
• Reallocating displaced GAA users to alternative channels, if available, while maintaining the aggregate interference budget

The CxM processes measurement reports from Citizens Broadband Radio Service Devices (CBSDs) to build a real-time interference map. This data includes:

• Received Signal Strength Indicator (RSSI) from neighboring cells
• Geolocation coordinates and antenna height
• Operational bandwidth and transmit power This empirical data supplements propagation models to validate that theoretical protection is achieved in practice.

The CxM organizes GAA users into logical Coexistence Groups (CxGs) —collections of CBSDs that must be assigned mutually orthogonal resources. The manager:

• Registers and authenticates group membership
• Treats each group as a single interference entity for external coordination
• Allocates a contiguous block of spectrum to the group, leaving internal scheduling to a Group Management Protocol (GMP)

A decentralized optimization paradigm where each cognitive radio acts as an autonomous learning agent. Instead of a centralized CxM dictating every channel, MARL allows agents to learn a Nash Equilibrium for channel access through trial and error.

• Agents receive a reward signal for successful, low-interference transmissions.
• Eliminates the single point of failure inherent in a purely centralized CxM.
• Often uses Deep Q-Networks (DQN) to handle high-dimensional state spaces.

A deep learning architecture that models the wireless network as a dynamic graph. Transceivers are represented as nodes, and the potential for harmful interference between them is represented as weighted edges.

• The CxM uses a trained GNN to predict aggregate interference margins in milliseconds.
• Scales efficiently to dense deployments where traditional matrix calculations are too slow.
• Enables the CxM to solve the Distributed Constraint Optimization (DCOP) problem for channel assignment.

The mathematical framework for formalizing the CxM's core task: finding a globally optimal channel assignment that satisfies local interference constraints. Each access point is an agent with a variable (its channel) and a constraint (cannot interfere with a neighbor).

• Algorithms like Max-Sum or ADOPT are used for distributed solving.
• The CxM acts as the coordinator, aggregating local utility functions to maximize proportional fairness.
• Avoids the computational explosion of a purely brute-force centralized search.

A multi-dimensional, real-time geospatial database that serves as the CxM's situational awareness layer. It integrates live sensor data, propagation models, and regulatory policies.

• Provides the CxM with the estimated power spectral density at any coordinate.
• Enables proactive spectrum mobility by predicting where a primary user will appear next.
• Constructed using Kriging or deep learning interpolation of sparse sensor measurements.

The process executed when the CxM orders a secondary user to vacate a channel. The goal is a seamless transition with minimal latency and zero packet loss.

• Proactive Handoff: The CxM pre-assigns a backup channel based on predicted occupancy.
• Reactive Handoff: The device must instantly search for a new hole upon a suspension order.
• Requires tight integration between the CxM's control plane and the device's MAC-layer scheduler.