Dynamic Spectrum Access Decisions. George F. Elmasry

tradeoff between the probability of false alarm and the probability of misdetection.

      This book is not about developing cognitive engines. However, some critical threads of this cognitive engine skeleton are worth covering to show some key aspects of the decision‐making process.

      5.2.1 The Main Thread in the Central Arbitrator DSA Cognitive Engine

This section shows how the generic cognitive engine construct used in the central arbitrator can execute the thread illustrated in Figure 5.3. The importance of this thread is that it is the initialization thread and subsequent threads can be considered a subset of this thread, as indicated by the curved arrow at the right‐hand side of the figure. The curved arrow means that the update to the information repository, the fusion of information, and the monitor of resources will continue to occur as a loop. At any moment, this loop can trigger a decision to propagate a new policy, rule set or configuration parameters or propagate a new frequency assignments. Note that there is no need for a human in the loop. The central arbitrator decisions are autonomous. Configuration parameters can be updated and propagated autonomously.

Figure 5.3 The main thread in the central arbitrator.

      Notice with Figure 5.3 that information repository updates can be:

       from external sources to include sensors, external database, and external tools.

       geolocation information updates propagated from lower hierarchy.

       a result of fusion that evolves the states of the DSA cognitive engine.

The decision maker can create a new frequency plan that includes:

       new frequency assignments to some or all networks (change of operational frequencies).

       new backup frequencies for each network (Note that backup frequencies, much like operational frequencies have to be a result of decision fusion so that they are spatially separated and when nodes use them for connectivity, nodes will not suffer from interference due to lack of spatial separation or from external systems in the same geographical location.).

       sensing frequency. (Note that in order to reduce the control traffic volume, the central arbitrator may select a subset of sensing capable nodes per each network. Not all sensors must be activated at all the times. This subset of sensors can be geographically dispersed to create a comprehensive spectrum map for the least amount of spectrum sensing control traffic volume.)

      5.2.2 A Critical Thread in the Gateway DSA Cognitive Engine

      One critical trigger that can reach the DSA cognitive engine in the gateway node is the receipt of a frequency plan from the central arbitrator. As mentioned earlier, the gateway DSA engine can defer frequency change in its network to the central arbitrator or the central arbitrator's own fusion can trigger a frequency change (frequency plan) that affects one or more networks. This frequency plan can contain frequency change for a network that the gateway is a node of, a list of frequencies to sense within the network identifying which sensor would sense which frequency, and a list of backup frequencies for the network to consider when network formation using the operational frequency fails. This is one of many possible paths that can go through the gateway DSA cognitive engine.

As Figure 5.4 shows, once the frequency plan is received by the gateway cognitive engine, it will result in an update to the gateway information repository and this update should trigger the decision maker to perform the following steps:

      1 Disseminate frequency change to all nodes in the network with a frequency change time that is dependent on the waveform characteristics. For example, if the waveform uses TDMA with EPOCH time, the frequency change time defined by the gateway DSA engine can be at the start of the next EPOCH or it can be after N EPOCHs from the current time. It is critical to ensure that the frequency change message will reach all nodes before the actual change of frequency can occur and that all nodes synchronously switch to the new frequency.5

      2 Schedule for a frequency change locally to the gateway node so that the gateway and all the nodes in the network can continue the same network formation with the new frequency.

      3 If the frequency change message is for a network that uses cooperative distributed DSA techniques, the frequency change message will contain a new pool of frequency bands for the network. The gateway DSA engine will have more work to do as it will have to disseminate the new pool and continue to perform cooperative distributed DSA within the network so that all the nodes in the network collectively try to optimize the use of the new pool of frequencies.

Figure 5.4 Gateway DSA cognitive engine thread for a frequency change.

      Note that Figure 5.4 is one possible thread the gateway DSA cognitive engine may execute. This is an example of a thread that propagates down the hierarchy. The next section gives an example of a thread that propagates up the hierarchy.

      5.2.3 The Gateway Cognitive Engine Propagation of Fused Information to the Central Arbitrator Thread

      This is an upward thread that can occur in the gateway DSA cognitive engine. With this thread, the gateway can be receiving spectrum sensing information from the following sources:

      1 Local, if the gateway has augmented spectrum sensors or the gateway can be sensing the frequencies in use as explained in the previous chapters.

      2 From peer gateway nodes. As explained earlier in this chapter, gateway nodes can be sharing their fused spectrum sensing information with peer gateway nodes.

      3 From the network nodes. The gateway node can be a member node of one or more networks. These networks can have augmented sensors configured to sense (probe) certain frequency bands and can be forwarding spectrum sensing information of the used frequencies to the gateway.

Figure 5.5 shows the progress of this thread to the fusion part. Collected spectrum sensing information from the different sources are fused to produce actions such as the following:

      1 Spectrum awareness map. This map shows what frequency bands in the area of operation of this network suffer from interference, what probed frequency bands can be used, and what probed frequency bands should be avoided because they are occupied. As explained earlier, this spectrum awareness map at the gateway is more accurate than the spectrum awareness map at the network nodes and the spectrum awareness map can be more accurate if peer gateways share spectrum sensing information with each other in a distributed cooperative manner.

      2 Reprioritizing of backup frequencies. The probing of unused frequency within the network can lead to changes in the backup frequency order of use. Also for a network that uses cooperative distributed DSA internal to the net, spatial use of the assigned frequency pool can be altered with time.