RC G1 Tactile Glove — Data Protocol & Pairing Reference

Use this page when you’re writing your own reader or integrating the G1 into a custom data pipeline. It covers the binary frame format, the AT command set on the serial channel, every LED state, and the most common failure modes. If you only need to plug in and record, start with the Windows, macOS, or Linux setup pages.

1. Hardware Architecture

A G1 deployment has five components. Only the first four are required for a single-glove rig; the USB hub is for multi-glove (LH + RH) setups.

1. Fabric sensor array. 162 piezoresistive sensing points arranged on a 16×16 logical matrix (256 ADC slots, of which 162 are physically wired to sensors; the unwired slots read near zero and can be masked in software).
2. Data-acquisition module (A-board). ESP32-based Bluetooth MCU with an integrated 300 mAh / 3.7 V battery, USB-C for charge / data / firmware, a hardware tactile button, and a status LED. 50 × 32 × 15.5 mm; 6.6 g board + ~15 g battery.
3. Bluetooth USB receiver (dongle). 1:1 paired with a single A-board. USB-A, 66.1 × 23.2 × 13 mm. Optional — cable-only operation is fully supported.
4. USB 3.0-to-Type-C cable (2 m). Serial connection at 921 600 bps default, optional 3 000 000 bps high-speed mode.
5. USB hub. For multi-glove rigs (e.g. bimanual LH + RH).

Data flow:

  +--------------------+     +--------------------+        +--------+
  |  Fabric sensors    | --> |  A-board ADC       | --+--> | Host   |
  |  (162 / 16x16)     |     |  (ESP32 + IMU)     |   |    |        |
  +--------------------+     +--------------------+   |    +--------+
                                       |              |
                                       | BLE 30 Hz    | USB-C 100 Hz
                                       v              | (921600 / 3M)
                              +--------------------+  |
                              |  BT USB dongle     | -+
                              |  (CH340 endpoint)  |
                              +--------------------+

2. Serial Configuration

• Default port settings: 921 600 baud, 8N1, no flow control.
• Optional high-speed mode: 3 000 000 baud — requires a firmware-side config change, contact support.
• USB chip on both the data-acquisition module (when wired) and the Bluetooth dongle: CH340 (WCH USB ID 1a86:7523).
• Linux: in-tree kernel driver ch341. macOS: install CH34xVCPDriver.app. Windows: install CH341SER.EXE.

The same binary protocol described in §3 works on either the wired or the BT-dongle serial path — you should not need to special-case the transport in your reader.

3. Binary Frame Format

A complete sensor reading is delivered as two packets that together carry 256 pressure values plus 16 bytes of IMU data. Each packet begins with the constant 4-byte sync header AA 55 03 99, followed by a 1-byte sequence number (0x01 or 0x02) and a 1-byte sensor-type ID.

Sensor-type IDs

Worked hex capture

Raw bytes off the wire from a live left-hand device sitting on a desk (truncated for readability):

[14:04:29.427]  AA 55 03 99 01 01 00 00 00 00 00 00 00 00 00 00 ...  (128 B payload)
[14:04:29.428]  AA 55 03 99 02 01 00 00 ... 00 62 F3 00 00 00 ...      (144 B payload, last 16 B = IMU quaternion)

Reference parser (Python)

Minimal parser that takes the two raw packets and returns a sensor ID, the 256-cell pressure vector, and the quaternion. Mirror of what the Linux setup guide uses.

import struct, numpy as np

def parse_pair(packet1: bytes, packet2: bytes):
    assert packet1[:4] == packet2[:4] == b"\xAA\x55\x03\x99"
    assert packet1[4] == 0x01 and packet2[4] == 0x02
    sensor_id = packet1[5]                              # same on both halves
    pressure = np.frombuffer(packet1[6:134] + packet2[6:134], dtype=np.uint8)  # 256 pts
    qw, qx, qy, qz = struct.unpack("<ffff", packet2[134:150])
    return sensor_id, pressure, (qw, qx, qy, qz)

4. Logical-channel map

The 162 wired sensing points are addressed in a 16×16 grid (256 cells, indexed row-major 1–256). The exact wired-cell list ships in ~/Documents/JQTools/docs/channel-map.csv with the companion app. Use it to mask out the 94 unwired cells when computing per-finger averages.

Sample rate: 100 Hz wired or 30 Hz Bluetooth. Single-point dynamic range: 0–350 N, resolution 0.01 N.

5. AT Command Set (Bluetooth pairing)

Send these as ASCII strings with a \r\n line ending on the BT dongle’s serial port at 921 600 baud. The dongle responds in ASCII, then switches to binary sensor data once a connection is established.

Worked example

> AT+SCAN=1
< OK
< +SCAN=1, 8277169C745A, -1, 18, Redmi Watch 3 745A
< +SCAN=2, 3C8A1F2E9A36, -1, 23, JQ-RH_3C:8A:1F:2E:9A:36   ← the device we want
< +SCAN=3, 0,88F3079C11B4, -1,9, KeepB4-B4

> AT+CONN=3C8A1F2E9A361
< OK
< (binary sensor frames begin streaming ...)

6. LED State Reference

Data-acquisition module — button functions

Data-acquisition module — status LED color

Bluetooth dongle LED

7. Re-pairing (if a glove “lost” its dongle)

1. Power-cycle the data-acquisition module: long-press until red, then short-press to advertise (blue flashing).
2. Plug the Bluetooth dongle into your host. Red flashing on the dongle means it’s scanning.
3. Open the host’s serial port to the dongle at 921 600 baud — use screen, minicom, the SSCOM utility shipped on the USB drive, or the Companion App.
4. Run AT+SCAN=1. Copy the MAC from the line ending in JQ-LH / JQ-RH / etc.
5. Run AT+CONN=<that MAC>. OK followed by binary frames means you’re paired. The link is persistent — you don’t have to re-pair after future power-cycles unless you swap modules or dongles.

8. Troubleshooting cheat-sheet

9. Specifications snapshot

10. Related pages