triangulation master thread

[legacv]

Distance = 10 ^ ((MP – RSSI)/(10 * N)) from here
useful stackoverflow article about distance from beacons

todo:

• find our MP = measured power - this is RSSI @ 1 meter. need to do a couple dozen(?) tests to get a good estimate
• find our N = environmental factor = how much noise is going to be fucking with us. ranges from 2-4

[dhanishpatil]

3 fixed BLE beacon
A: (x₁, y₁)
B: (x₂, y₂)
C: (x₃, y₃)

1 handheld device

RSSI = Tx power(RSSI value when the handheld is 1 feet away) - 10nlog(d)

Find tx power
Place your receiver exactly 1 meter from the beacon.

Record RSSI (take average of ~20 readings).
→ That average = TxPower.
Find n

Move the receiver to known distances (e.g., 2m, 3m, 5m).

Record average RSSI at each distance.

Fit the formula for n:
$n=\frac{(TxPower-RSSI)}{10log_{10}(d)}$

Example

Find d

$d=10\times\frac{TxPower-RSSI}{10n}$

So we will have da, db, and dc values and (xa,ya) (xb,yb) and (xc,yc) coordinates and we need to find x,y coordinates for the hand held

When we have multiple beacons senders we will only select 3 positions that have the strongest rssi value.

Make sure to pick 3 coordinates where at least the the middle one should be on the opposite side
=> Solution: Like if the 1 and 3 is on xa,ya and xb,yb then the second should have a different x value and should be between ya and yb

Use latest rssi values to calculate d and then calculate x,y

Once every second we calculate the distance

Have 3 rssi values from different beacon (sensor) before calculation

[CyberSamurai0]

I had an idea regarding energy conservation down the line. Since the handheld system is intended to send a BLE response, we can use that response to trigger a "wake-up" and send more frequent advertisements when a device is nearby.

Algorithm Breakdown

Default Behavior:

1. Deep sleep 3 seconds
2. Send BLE advertisement
3. Scan for response
4. If we receive a response to our advertisement:
   1. At least one device is nearby
   2. Switch to active behavior
5. Repeat

Active Behavior:

1. Deep sleep 500ms (or 100ms, 300ms, etc.)
2. Send BLE advertisement
3. Scan for response
4. If no response:
   1. Increment silent counter
   2. If silent counter hits threshold (a few seconds worth)
      1. Switch to default behavior
5. Else if response:
   1. Clear silent counter
6. Repeat

[CyberSamurai0]

Now that we have working advertisement payloads, RSSI seems to be particularly unreliable, so we might have a precision level falling within a matter of meters. However, we should still be able to get decent accuracy if we have enough nodes in an area.

True triangulation (angle finding) requires that the handheld have a multi-input antenna array and use BLE 5.1. We had planned to be backwards compatible with BLE 4.0, so we may have to stick to RSS-based trilateration.