PE1-5.5 Cognitive Radio & Spectrum Sensing

What this is and why it exists

Cognitive radio is software defined radio that decides for itself. It observes the spectrum around it, chooses where to transmit, and adapts.

The motivation is an observation anyone can verify with a spectrum analyser. Licensed spectrum sits mostly idle in time and in place, while unlicensed bands are crowded.

The vocabulary

• Primary user — a licensed holder with the right to transmit in a band.
• Secondary user — a cognitive radio using the band when the primary user is not.
• Spectrum hole — a band unused at a particular time and place.
• Cognitive cycle — the loop of observing, deciding, acting and learning.
• Spectrum sensing — determining whether a band is in use.
• Energy detection — deciding a band is occupied when the measured power exceeds a threshold.
• Hidden node — a transmitter that one sensing radio cannot hear but another can.
• Cooperative sensing — several radios combining their observations into one decision.
• Reporting channel — the bandwidth spent sending those observations to be combined.

The mental model

The motivating fact deserves stating carefully. Spectrum is allocated administratively and used sparsely. A band licensed to a broadcaster is occupied where that broadcaster transmits and idle elsewhere, and often idle there too at some hours. Meanwhile the unlicensed bands everyone shares are congested. The resource is not scarce; the allocation is rigid.

A cognitive radio uses those gaps. It senses whether a licensed user is present, transmits when one is not, and stops when one appears. The licensed holder keeps priority throughout, which is what makes the arrangement acceptable to a regulator at all.

That last point is why this is a policy problem as much as an engineering one. The rules governing who may transmit were written for fixed allocations, and they would have to change for this to be permitted at scale. That is unusual in an engineering subject and worth noticing.

The cognitive cycle is the organising idea. Observe the environment. Decide what to do. Act on that decision. Learn from what happened, so the next decision is better. Every capability in this topic sits at one point on that loop.

Architecturally, such a radio is an ordinary chain with sensing and decision added. The sensing and the decision must both change with circumstances, so a software radio is the natural base for it. That is what joins the two halves of this subject.

Energy detection is the simplest sensing method and its weakness is instructive. Measure the power in a band and compare it against a threshold. If the power is high, assume someone is transmitting.

It needs no knowledge of the signal, which is its strength. It also cannot tell a signal from a rise in the noise, which is its weakness. At low signal-to-noise ratio the two are indistinguishable. It needs to know the noise power to set its threshold. If that estimate is wrong, the detector is wrong in a way it cannot notice. Under fading a real transmitter may momentarily look absent.

The hidden node problem is worse than any of that. A cognitive radio behind a building may be unable to hear a licensed transmitter that is nonetheless being received on the far side. It concludes the band is free, transmits, and interferes with exactly the user it was meant to protect.

Cooperative sensing fixes that. Several radios in different places sense the band and combine their observations, so a transmitter hidden from one is heard by another. The combining can be as simple as declaring the band busy if any node reports it busy.

The challenges are the interesting part. Reports have to be sent somewhere to be combined, and that reporting channel costs bandwidth that grows with the number of nodes. The nodes must agree on when they sensed, or they are combining observations of different moments. And a node that reports dishonestly, whether faulty or malicious, can poison the shared decision. A single node claiming a band is free can cause every other radio to interfere.

The course closes on where the field is heading. Read that part for direction rather than detail, because it is the material most likely to have dated.

What you should now be able to explain or do

State the observation that motivates cognitive radio and why priority stays with the licensed user. Describe the cognitive cycle and place each capability on it. Say why this is partly a policy problem. Explain energy detection, its assumption about noise power, and the two conditions under which it fails. Explain the hidden node problem, how cooperative sensing solves it, and the three costs that come with it.

Check yourself