Wechat small program iBeacon ranging and stabilization program

Once you know rSSI and txPower, you can calculate the distance. There are two formulas:

A,

The three variables A, B and C in this formula are all empirical values, which need to be accurately adjusted according to the mobile phone system or hardware model. Usually, the calibration results of all devices are saved into A device information table. The mobile terminal first checks the model of the machine, and then obtains the corresponding calculation configuration based on the device information. Obviously, this formula is more dependent on hardware tuning, and it will be difficult to use this formula without data storage.

Convert to js code:

Const calculateAccuracy = function (txPower, rssi) {return (0.89976) * math.pow (rssi/txPower, 7.7095) + 0.111}Copy the code

Distance measurement performance under the condition of not fine calibration:

So let’s talk about what this graph looks like.

The vertical axis represents the measured distance, and the horizontal axis represents the time. Samples are taken every second. The figure shows the numerical curve of nearly 10 samples. The green line is the RSSI value received by the device, reflecting the actual data received by the hardware. The red line is the instantaneous distance calculated by the formula; The yellow line is the instantaneous ranging result of wechat applets.

The actual distance between bluetooth beacon and mobile phone is 1m, and the test device is Redmi Note7.

As can be seen from the figure above, the RSSI value is relatively stable, indicating that there is no major problem with the hardware. The red line and the yellow line both fluctuate wildly, which means they’re not very accurate. The fluctuation trend of the two is almost the same, so there is reason to suspect that the micro channel small program is also used in the ranging formula. According to the results, the accuracy of this formula is relatively poor, which may be due to the lack of precision correction.

Second,

The A in this formula is RSSI, tx is txPower, n is the empirical value, and n depends on the physical environment.

Convert to js code:

const calculateAccuracy = function (txPower, rssi) {
  return Math.pow(10, Math.abs(rssi - txPower) / (10 * 4))
}
Copy the code

Range finding performance of Formula 2:

More people die than people, more goods to throw away.

The yellow line in the figure is still as crazy as the fluctuation, but the red line is unusually stable, and shows the same fluctuation range as the green line, indicating that the ranging accuracy is reliable. This formula has only one parameter, and the adjustment in the production environment is relatively simple. Here, we choose Formula 2 as the ranging formula.

The stabilization program mainly does “peak clipping and valley filling” for the band data, also known as data filtering. The simplest filter processing, is the collection value and the average period of time, as long as the hardware is not a problem, fixed distance bluetooth beacon rssi values will always be in a relatively stable range of change, the longer the sampling time, the average of the sample will be more close to the real value, so in the static range scenario, the average is the best way.

Const arrayAverage = arr => arr.reduce((acc, val) => acc + val, 0)/arr.length; return arrayAverage([...] )Copy the code

The specific realization is that when the program continuously receives the RSSI of the beacon, it first calculates the instantaneous ranging result with the formula, and then saves the result into an array, and then calculates the average value of the array. Static ranging, the measurement results are very accurate, the distance error within 2m can be as low as 0.1m.

In practical applications, it is often dynamic ranging, so the length of sampling data should be limited, such as taking the latest 10 groups of data in the last in first out order. The specific sampling queue length depends on the actual requirements of the project. The longer the sampling queue is, the smoother the ranging result is, but the duller the dynamic capture of mobile terminal is. On the contrary, the shorter the sampling queue, the sharper the result, the more sensitive to the dynamic capture of mobile terminal.

Sometimes because of some accidental factors, sampling the queue will appear the “noise” of individual substantially deviates from the value of the real data, is difficult to effectively erase effect even for the average, to eliminate the effect, can be in average before using gaussian fuzzy algorithm of “large value” and “small” value, smoothing the maximum noise reduction.

The key of Gaussian blur algorithm is to calculate the weight according to the average difference. The weight calculation formula of one-dimensional Gaussian blur is as follows:

Convert to js code:

// Find the gaussian fuzzy weight @param(queue length, Const getOneGuassionArray = function (size, kerR, sigma) {if (size % 2 > 0) {size -= 1} if (! size) { return [] } if (kerR > size-1){ return [] } let sum = 0; let arr = new Array(size); for (let i = 0; i < size; i++) { arr[i] = Math.exp(-((i - kerR) * (i - kerR)) / (2 * sigma * sigma)); sum += arr[i]; } return arr.map(e => e / sum); }Copy the code

The concepts of “large” and “small” are described below, but it is important to know that our vague target is “extreme data”.

After time weighted processing data lag will have certain inhibition, but if there are any more “steep” distance changes, this process will still be given a relatively “smooth” feedback, dynamic change in order to make stable program can better awareness, follow up and respond, also need to set up some special conditions.

First, how do you tell if a mobile device is moving away or closer?

Here is a simple idea to find the maximum and minimum values in the sample queue, and then find the partial large and partial small values at a certain threshold. For example, if the maximum value of the queue is 3, the minimum value is 1, and the threshold value is set to 0.1m, the data larger than 2.9m is considered to be larger than the maximum value, and the data smaller than 1.1 M is considered to be smaller than the minimum value. The maximum length of the larger and smaller queues should not exceed half of the total queue.

Then, if the larger values are concentrated in the first third of the queue, we can assume that mobile devices are decisively moving away; Otherwise, the smaller values are concentrated in the first third of the queue, which can be considered as mobile devices are approaching.

If (arrayAverage(maxCount) < parseInt(queueLength / 3)) {//queueLength < parseInt(queueLength / 3) { Console. log(' moving away ')} else if (arrayAverage(minCount) < parseInt(queueLength / 3)) {console.log(' moving away ')}Copy the code

Based on this tendency to move away and closer, we can artificially tilt the data more aggressively in the direction of motion. So how do we do that? Time weighting logic is skipped. If it is judged to be moving away, the smaller value in the queue will be filtered out; otherwise, the larger value will be filtered out and only the remaining data will be calculated. This would produce an obviously overzealous result, but given the inertia of movement in the real world, such a radical approach would probably be more in tune with real movement and make the data more “responsive”.

Bluetooth range is simply the application of a formula, itself is simpler, based on distance measurement can realize a lot of near field application, such as the near-field check-in, near-field push, etc., can even further realize the localization of mobile devices, with the location information, a lot of indoor navigation positioning and the related application can be done, The next article will explain in detail how to implement location based on iBeacon technology.

Front-end road original technical article, reproduced please indicate the source: refined-x.com/2019/03/30/…

Unwilling to mediocre you, come to charge with me, pay attention to the scenery, get more wonderful content.