Dynamic Spectrum Access (DSA)

The foundational capability of a DSA-enabled secondary user to autonomously monitor the electromagnetic environment. This involves detecting the presence or absence of Primary User (PU) signals to identify Spectrum Holes—frequency bands that are unoccupied at a specific time and location.

• Cyclostationary Feature Detection: Exploits the periodic statistical properties of modulated signals to distinguish them from noise, even at low SNR.
• Matched Filter Detection: Maximizes SNR when the secondary user has a priori knowledge of the primary user's signal structure.
• Energy Detection: A simpler, non-coherent method that measures the energy in a band but struggles to differentiate between a primary user and interference.
• Cooperative Sensing: Multiple radios share local observations to overcome the hidden node problem caused by shadowing and multipath fading.

The absolute regulatory and technical imperative of DSA. A secondary user must vacate a channel immediately upon detecting a returning primary user, a process known as Spectrum Handoff. This is governed by a strict interference temperature limit, which caps the cumulative RF energy a secondary user can introduce at a primary receiver.

• Listen-Before-Talk (LBT): A collision avoidance protocol requiring a clear channel assessment before transmission.
• Underlay Access: Secondary transmission is permitted concurrently with a primary user only if its power spectral density remains below the noise floor at the primary receiver.
• Overlay Access: A more complex cooperative technique where the secondary user uses advanced coding (e.g., dirty paper coding) to mitigate interference while potentially relaying primary user traffic.

The ability of a cognitive radio to seamlessly vacate its current operating frequency and reconfigure its transmission parameters to operate on a new target channel. This Spectrum Handoff must be executed with minimal latency to prevent service disruption.

• Proactive Handoff: The radio predicts the arrival of a primary user via Spectrum Occupancy Prediction models and switches channels before the incumbent appears, reducing collision probability.
• Reactive Handoff: The radio switches channels only after detecting a primary user, requiring a fast-sensing mechanism and a pre-defined backup channel list.
• Parameter Reconfiguration: Beyond frequency, the radio adjusts modulation, coding rate, and transmit power to match the characteristics of the new spectral environment.

Modern DSA systems move beyond reactive threshold-based logic by employing Reinforcement Learning (RL) agents. These agents learn optimal channel selection and access policies through interaction with the RF environment, modeled as a Partially Observable Markov Decision Process (POMDP).

• Deep Q-Networks (DQN): Approximate the action-value function to predict the long-term reward of selecting a specific channel, handling high-dimensional spectrum occupancy data.
• Multi-Armed Bandit (MAB): A simplified RL framework that balances the Exploration-Exploitation Trade-off—trying new channels versus sticking with known good ones—to maximize cumulative throughput.
• Multi-Agent RL (MARL): Coordinates multiple secondary users in a shared spectrum environment using Centralized Training Decentralized Execution (CTDE) to prevent collisions between secondary users themselves.

A regulatory mechanism that serves as an alternative or complement to spectrum sensing. Secondary users query a centralized Spectrum Access System (SAS) with their location to receive a list of available channels and permissible power levels. This is the primary access model for the Citizens Broadband Radio Service (CBRS) band.

• Three-Tier Hierarchy: Incumbent Access (highest priority, e.g., Navy radar), Priority Access License (PAL), and General Authorized Access (GAA).
• Propagation Modeling: The SAS uses terrain data and propagation models (e.g., Irregular Terrain Model) to calculate exclusion zones and protect incumbents from aggregate interference.
• Environmental Sensing Capability (ESC): A network of sensors that detect incumbent radar activity and instruct the SAS to dynamically reconfigure the spectrum, providing an additional layer of protection beyond the database.

DSA fundamentally exploits the fact that licensed spectrum is highly underutilized across time, frequency, and space. A Spectrum Hole is a three-dimensional opportunity defined by a specific frequency band, a geographic location, and a time window where no primary user is active.

• Temporal White Space: Gaps between primary user transmissions, such as the idle periods in TDD (Time Division Duplex) systems or radar rotation cycles.
• Spatial White Space: Geographic areas outside the primary user's coverage footprint, such as the guard zones between TV broadcast towers.
• Radio Environment Map (REM): A multi-domain database that integrates spectrum occupancy measurements, terrain data, and transmitter locations to build a comprehensive spatio-temporal model of spectrum usage for predictive access.