Cognitive radio, cognitive everything

Ron Wilson, Altera Corporation

Cognitive Radio (CR) has evolved from research themes to classified intelligence and military applications, as well as practical technologies for wireless networks. Today, it is about to launch more extensive applications: as the basic composition of each node of a large machine-to-machine network in the fields of transportation, utilities, building control, etc.; as a supporting technology for cellular femto base stations; and various applications other than communications.

For system designers, in addition to curiosity, there are two reasons that they are now starting to study CR. The first is that the concept of CR is increasingly used in system wireless networks, from chemical plant sensor networks to vehicle control and displays. To understand these systems need to understand the radio, this is because CR is not exactly the transmission network technology. The second reason is that CR's architecture represents the future of many other system-like architectures. You will also encounter these concepts.

Simply put, CR is a radio that can understand and adapt to its signal environment. This is an anthropomorphic explanation that requires further definition. The CR checks the power of its access spectrum, isolates the signals found in it, classifies them, compares the available bandwidth and the bandwidth currently required by the main radio service system, and then formulates and implements communication strategies to meet the system bitstream and delays. Claim. These different tasks have various different application focuses.

At present, cognitive concepts have been used in many different applications. The earliest applications were signal intelligence and electronic warfare. Here, CR can be used in at least three different modes. First, in intelligence applications, the CR examines a spectrum, detects the power that may contain signals, classifies these potential signals according to the modulation mode, possible sources, and levels of interest, and then configures the receiver channels for acquisition. Target transmission information. If you think that the most interesting content comes from trying to hide your own radio signals, these signals use spread spectrum short pulses, then you will see that in these applications, CR will continuously monitor a wide spectrum, or all recorded, Instantly lock the target signal.

Many electronic warfare applications are similar. For example, an interfering device will find and classify signals according to the methods described here, and then configure the transmitter channels to interfere with specific transmission information in real time. In battlefield communication applications, the CR scans and recognizes, and then places the information to be transmitted between other signals in the environment so that it is difficult to detect.

If military applications have already been carried out, civilian use will soon follow. The three fundamentally same functions are perception, classification, and adaptation that have been used in many other simple environments. For example, a cellular phone periodically checks for signals from nearby base stations, switches cells, and even switches air interface standards accordingly. Similarly, a cellular femto base station - the smallest base station you can buy, you can put it on the desktop while learning, this base station needs to continuously detect the data stream and its in-band interference level, depending on where and when The free spectrum in the range to adjust its power output, select the channel to use. In this way, the femto base station enhances the space utilization, thereby increasing the total network capacity without reducing capacity due to building interference. For example, Mindspeed Technologies has some key intellectual property in this area.

A new emerging application of CR is the mid-range machine-to-machine wireless network. Applications include smart grids, traffic control, and campus-level building management. According to William Webb, CTO of Weightless Special Interest Group, many of these networks use the so-called blank TV signal band: channels that are not occupied locally within the TV bandwidth. However, in this frequency band, no device is allowed to interfere with anyone’s television reception, nor can it interfere with a licensed blank TV signal radio, but it may interfere with other unlicensed low-power devices. In fact, Weightless's proposed solution is to recognize that wireless networks are based on small radio terminals that do not have cognitive capabilities.

The best way to understand how these various requirements affect radio design is to understand the capabilities of the device and to study how the application requirements change the design.

This type of discussion is best started with Software Defined Radio (SDR). According to Sofie Pollin, IMEC's ​​chief scientist, the ideal SDR has no analog function at all. This type of radio directly digitizes the entire frequency band from the antenna, then extracts the required signals and uses a software-programmed digital processor for demodulation. In this way, the software-defined parts: radio filtering, demodulation, baseband processing, and reverse processing on the sending side, can be performed in software, and can also be performed in hardware configured by software and parameter assignment. In this way, the radio equipment only needs to call different functions or transfer different parameters to change the channel, frequency band, modulation method, protocol, and error correction method.

Pollin quickly pointed out that this idea is difficult to achieve in practice. In many applications, the direct digitization of the antenna signal requires a very fast analog-to-digital converter (ADC) and a wide dynamic range, which itself is a scientific project. Therefore, the actual SDR is a compromise: a highly configurable analog RF stage, and a downconverter before the ADC can be configured, as shown in Figure 1.

Figure 1. The actual SDR receiver.

In this actual SDR, we added three functional modules to achieve radio cognitive function, as shown in Figure 2. These modules implement the three basic functions described earlier: perception, classification, and adaptation. How they change the SDR is a problem that is closely related to the application.

Figure 2. Cognitive radios add sensing, classification and adaptation modules to SDRs.

The first major module is awareness, which affects the front end of the receiver. Unlike SDR, CR must perceive its special environment. In less demanding applications, this sensing process may be as simple as quickly reviewing the reception of the programmable front end of the time domain multiplexing receive path. For example, the receiver scans a list of predefined channels between base station data packets.

Pollin pointed out: "The system for implementing the opportunity spectrum access (OSA) is very simple. Most of the emphasis is placed on the currently undisturbed frequency bands and then performing frequency hopping. This method is very powerful."

For scenarios that require fast response, the radio device contains a completely independent receive path and is only used to monitor the interval channel. Using a separate path, the CR can simultaneously process the defined frequency band and signal activity in the normal receive frequency band. In extreme cases, the CR will have completely different receive channels when implementing the sensing function. For example, a device will have an adjustable narrowband receive path for communication, and a very large bandwidth path for sensing so that it can continuously process a wider radio spectrum without scanning. Pollin said: "Depending on the requirements, the monitoring function may require a very large bandwidth analog front end." However, the cost of broadband circuits is very high, so you should try to avoid using it. If the actual application allows, you better use the scanning method.

The second important component of the SDR structure chart is the classification function. In the classification module, the CR processes the data stream collected by the sensing module, finds power patterns that may be interfering signals, or the power pattern of the target signal, and classifies them.

For a simple OSA radio, the implementation of the classification function is as simple as the establishment of a power histogram for the distribution channel, without concern for what power represents. No matter whether the channel is silent or signaled. More complex radio equipment can infer more energy sources, such as transmitter frequency, debug mode, and transmission timing: this is the information used by the strategy to form an adaptive module for making decisions.

James Neel, President of Cognitive Radio Technologies, explains: “Classification functions are generally implemented in two steps. First, you just probe for energy and then get more information.”

Sometimes it can be probed with statistical calculations to get more information. In other cases, the radio will try to tune the signal, demodulate it, and check the packet flow. In this case, the radio demodulator can be used for both the receiver and the classification purposes, or, in the case of front-end circuits, there is a completely independent digital processor for classifying the unknown signal. Neel pointed out, "If you know exactly what to look for, then you can design custom classification hardware."

If you don't know what to look for, then all the work will be fun. The categorization can begin by extracting the analog front-end data and then develop a power spectral density (PSD) function, such as through a Fast Fourier Transform (FFT). Then, from the power spectrum changes over time, the signal uses which modulation method and other aspects of the cycle analysis. Depending on the actual application requirements, these statistics can be obtained using digital signal processing (DSP) chips, FPGAs, or other computing hardware.

The classification module then transmits its result to the third unique component of the CR, ie, the adaptation module. Here, classification data and statistics face two key issues: What does it represent and what should I do? At this level, the CR architecture represents the biggest change in application drivers.

The difference lies in the scope of the problem. A simple OSA radio may simply observe its link interference level or bit error rate (BER) and look for open frequencies that can be frequency hopped. More complex radio equipment will construct the spectrum map of the interference source, predict how the spectrum map will change based on the observed behavior of the interference source over time, and plan a series of frequencies and modulation methods to meet its data communication requirements.

Neel said that in the most complex designs, CR not only knows its spectrum environment, but also knows where its system is located. For example, a CR in a hospital operating room can "understand" the surgical procedure performed, the status of the patient, the needs of each individual in the surgical team, and the behavior of other nearby wireless devices. This radio will combine all this information into a single strategy, using the available spectrum to guarantee the delay of emergency data flow while meeting the low priority requirements of other instruments.

Such radio devices may use heuristic functions to drive environmental analysis and make decisions. Neel suggested, or, as the environment becomes more and more complex, predictability becomes worse and worse. The strategy module will use hidden Markov models, neural networks, or related technologies to model the environment and analyze it.

This type of calculation involves a large amount of data and requires a strong calculation function. Neel said that, therefore, the obvious trend is to put CR's thinking part in the cloud. Then, a key issue is the duration of the phenomenon that the radio equipment has to deal with. Is there time to check the computing resources of the center, or do the radio devices have to make their own basic judgments?

With the popularity of CR, it has brought other problems. Inevitably, a large number of CRs cause spectrum congestion. They must seize each other's frequency band and provide their hosts with the highest bandwidth. As Neel pointed out, under these circumstances, radio equipment needs to use game theory in order to achieve optimal allocation of shared frequency bands. The problem of stability and optimization of this type of competing network brings a problem that is difficult to solve in the agenda of the IEEE 802.19 wireless coexistence technical advisory group.

Two examples will help to clarify these concepts. First of all, the Weightless standard is actually not a CR, but a cognitive network where very simple radios are just one component. The second is a theoretical example that applies to public safety applications with more complex true CRs.

The Weightless architecture is similar to IEEE 802.22 and uses a regional network within the so-called blank television signal band, which is an unoccupied local channel within the television broadcast bandwidth. Unlike 802.22, which is designed to provide broadband access to people, Weightless is specifically designed to provide narrow-band IoT connectivity for different machines, sensors, and actuators, with very low node and power consumption.

According to Weightless's Webb, different purposes lead to significant architectural differences. Webb said: “For devices that are not licensed in the blank television signal band, the first restriction is that it cannot interfere with anyone’s TV reception. This is hard to do because you know exactly where all the TV transmitters are located, but they’re not Know where the receiver is.

Webb continues to explain: "Our initial idea was to design a very low-cost CR for the terminal, which is very similar to the purpose of 802.22. However, we found that in reality, you may be close to the person who is watching TV, the signal emitted by a long-distance TV station. It was very weak when sent to him."

CRs with sufficient dynamic range to identify these weak signals require high costs and provide large power. Therefore, Weightless chose a completely different approach to perception: a three-tier network controlled by the center, as shown in Figure 3. Here, the terminal reports its physical location to the base station, the base station queries the application program in the cloud, and the cloud uses the location list of the licensed broadcast transmitter and other base stations to obtain a list of available frequencies. Webb commented: "This is an interesting algorithm for design."

Figure 3: The Weightless network spreads cognitive functions into three layers.

This look-up table ensures that the terminal will not affect everyone watching TV at night, nor will they compete with other base stations for the same channel. However, it does not guarantee that there will be no interference between the terminal and the licensed device, such as wireless microphones, or other unlicensed low-power devices that are roaming in the same area. To solve this problem, the base station allocates 8 frequencies to the terminal. The terminal performs frequency hopping at two-second intervals to reduce interference from such sources.

Because the terminal uses direct spread spectrum transmission technology, the system has greater freedom. In practice, the base station monitors the BER of each terminal and instructs the terminal to adjust the spreading factor in order to achieve a balance in bandwidth and range.

Weightless uses three defined behaviors of the cognitive system: perception, classification and adaptation. However, this standard does not build all these functions into the terminal radio equipment, but spreads them out. Most of the recognition is in quasi-static tables, not in algorithms. The adaptive function, most of the sensing functions and the classification functions are balanced in the base station. The distribution of functions in the network greatly reduces the cost of the terminal. The single-chip battery-powered device can work with a very low duty cycle.

The second example is not based on existing standards, or related actual radio, but based on the concept proposed by the Federal Communications Commission's Office of Public Safety and Homeland Security. However, a critical and long-standing problem faced by emergency responders in the United States is that different units cannot communicate with other units’ equipment via radio in emergency situations, such as police, firefighting, and medical teams.

Sometimes the problem is the availability of spectrum. For example, a recent public report showed that the 850 MHz band of the newly equipped police radio system in Oakland, California, is close to the local 2G cellular service, and communication errors often occur.

Other issues arise from their own perceptions. For example, there is a conflict between regular broadcast and official emergency communication bandwidth in emergencies. Other issues involve compatibility between different services. Police, fire, medical, and urban radio systems are not interoperable in an emergency. Bureau chief engineer Bill Lane and electronic engineer Yoon Chang wrote in their technical topic article “Public Safety for Cognitive Radio” that as narrowband voice radios become media terminals similar to high-performance smart phones, these problems will only increase. The more serious. Lane recommends using CR as a solution. The CR can find available spectrum in real time near the emergency point, avoiding interference sources. It can use the available spectrum preferentially to guarantee the lowest critical message delay. By finding the frequency, modulation and coding of the surrounding radio equipment, the CR can bridge incompatible systems in the area.

A radio device that can complete all of these services will require that all three cognitive modules be implemented locally: perceptive, classified, and adaptive, as shown in Figure 4. For sensing, the radio needs a dedicated monitoring chain. The good news is that this chain does not require a very high bandwidth, ultra-high dynamic range receiver. For applications that need to scan receiver channels in the target bandwidth, a set of data is built for each channel.

Figure 4. Cognitive Radio of Emergency Response Personnel.

Classification hardware is much more complex than simple OSA radios. The FFT stage and multiplier will create a PSD function for each frequency band. Then, the signal processing chain will apply a periodic analysis to extract the characteristics of a certain modulation method. In this way, the classifier will provide the adaptive module with a large amount of information for each frequency band: where the interference is, where the signal can be identified, and enough statistics to find other nearby radio signals.

Process this data and combine it with pre- and post-process information that meets application requirements, user priorities, and meets interoperability requirements. In multi-core CPU clusters, an adaptive module implemented in software uses neural network simulation to find open-frequency-specific Radio type and time domain mode. Using more rules, the software will choose a strategy to program the analog front end and base station part of the radio to implement this strategy.

These examples illustrate the use of cognitive technology in data communications. And these concepts apply not only to radios. In fact, when the system is in a rapidly changing environment, not only the parameters need to change, it is best to design the system response as a cognitive system.

Perceived, classified, and adaptive as the three components of the link function have been widely used in intelligence intelligence and electronic warfare. In many other areas, the effective application of these concepts will generally have low costs, including network management, active network security, advanced car-assisted driving, and industrial control.

Current factors include rapidly increasing density, and lower and lower power budgets and computational costs. In the near future, most electronic systems will be cognitive systems, with fewer and fewer fixed-function systems, and fewer special new designs. This development trend will have a profound effect on the actual system design.

Watch For Girls

Jinhu Weibao Trading Co., Ltd , https://www.weibaoe-cigarette.com