Wireless communication is facing several challenges, including spectrum scarcity, deployment delays, and formation and management of dynamic networks. Cognitive radios (CRs) will help address these challenges, but only if they reason about policies to guide their behavior. Policies are sets of declarative statements with unambiguous semantics. At the highest level of abstraction, an XG (neXt Generation) radio has four main components, as shown in the illustration below;
1. Sensors. XG radios need sensors to discover available spectrum and transmission opportunities.
2. Radio Frequency (RF). The RF component transmits and receives.
3. System Strategy Engine (SSE). The SSE controls the radio’s transmissions, and can transmit only when the PE has approved transmission. It builds transmission requests based on sensor data received from the sensors and its current strategies. Replies to its requests from the PE may affect strategy.
4. Policy Engine (PE). The PE accepts transmission requests from the SSE and checks policy conformance. The PE has all active policies loaded. The PE is platform independent. The PE runs on the radio—as opposed to other possible architectures, including distributed models—and it must be compiled for the target hardware.
Several different types of messages can be sent between the different components of the architecture:
RF−SSE (system strategy engine). All incoming messages to the XG radio arrive at the RF unit, and end up in the SSE. These can be control messages (e.g., updates to system strategies, updates to policies, or messages controlling the coordination with other radios). Similarly, all messages going out from the XG radio originate in the SSE and are passed through the RF component. Outgoing messages can also be control messages or data messages.
Sensors−SSE. The details of this interface will be determined by the radio designer. We assume that the sensors send their received data (or conclusions drawn from it) to the SSE. The analysis of sensor data, sensor data aggregation, signal detection, and other such processing could happen in the sensor component(s), in the SSE, or in a dedicated component (not shown). The SSE may send control messages to the sensor components.
SSE−PE. Several types of messages exist in the interface between the SSE and the PE.
■ Transmission requests: The SSE builds a transmission request, and sends it to the PE. The PE reasons about the request and the active policies, and responds by sending one of three types of replies: (1) the transmission is allowed, (2) the transmission is not allowed, or (3) specified additional constraints must be satisfied. Given acceptable values of the underspecified request parameters, the transmission will be allowed.
■ Policy updates: The SSE can send policy-update messages to the PE, to update the PE’s policy base, by adding or removing policies and activating or deactivating policies.
■ Policy information: The SSE can request information regarding which policies are loaded or active. The PE never initiates a message exchange. It should be noted that this interface minimizes the amount of “state information” that the PE needs to keep. The only persistent state is the set of loaded policies. Other than that, each message and reply is independent from previous messages and replies.
Given the open-ended nature of the SSE−PE interaction, how should the SSE behave? What should it request? The XG architecture does not offer any answers to these questions. A broad range of radio devices, some of which may have a sophisticated SSE component, and others that may have a simpler SSE are envisioned.
A simple SSE might put only the requested frequency and maximum power in the request and hope for approval. If additional constraints are returned, it might just try another frequency from a table until the request is approved—or the SSE may just give up. A cognitive SSE, on the other hand, might exploit spectrum opportunities by constructing a request that satisfies the additional constraints returned by the PE. For example, the SSE might have to perform sensing actions and include information about the sensed spectrum in its requests in order for them to succeed.
