How to Build a Multistatic Mesh with ESP32 for 360-Degree Room Coverage Using RuView

You can build a 360-degree multistatic sensing mesh by deploying 4–6 ESP32-S3 nodes running RuView's CSI firmware, provisioning them with unique node IDs, and running the Rust-based RuView server to fuse multistatic Channel State Information (CSI) into full-room pose estimates.

RuView is an open-source Wi-Fi sensing stack that transforms inexpensive ESP32-S3 hardware into a professional-grade multistatic radar. By following the ADR-029 – RuvSense multistatic sensing mode specification, you can achieve sub-inch pose jitter, vital-sign monitoring, and persistent-field modeling across an entire room using only Wi-Fi signals.

Hardware Layout for 360-Degree Coverage

For complete room coverage, position nodes to maximize TX-RX link diversity. According to ADR-029, the baseline configuration uses 4 nodes placed at wall midpoints approximately 2 meters apart, creating roughly 12 unique links. For production-grade robustness, deploy 6 nodes to generate 30 redundant links (N·(N-1)), improving SNR and enabling multi-person separation.

Each node runs identical firmware from firmware/esp32-csi-node/main/csi_collector.c, which handles UDP CSI serialization, channel hopping, and rate-limited output to prevent lwIP buffer exhaustion.

Flashing the ESP32 CSI Firmware

Build the firmware using the provided install script, or flash the pre-built binary directly.

To build from source:

./install.sh --profile iot --yes

This generates firmware/esp32-csi-node/build/esp32_csi_node.bin.

To flash a node (replace port as needed):

esptool.py --chip esp32s3 --port /dev/ttyUSB0 \
  --baud 460800 write_flash 0x1000 \
  firmware/esp32-csi-node/build/esp32_csi_node.bin

The firmware implements channel hopping across 2.4 GHz channels 1, 6, and 11 (configurable in csi_collector.c lines 66-73) and a TDM schedule where each node transmits null-data-packets in assigned slots while others listen, as defined in rust-port/wifi-densepose-rs/crates/wifi-densepose-hardware/src/esp32/tdm.rs.

Provisioning Each Node with Wi-Fi Credentials

Use scripts/provision.py to write Wi-Fi credentials, node IDs, and sensing parameters into the ESP32 NVS partition. Each node requires a unique --node-id (0-255) so the server can differentiate streams.

python scripts/provision.py \
  --port /dev/ttyUSB0 \
  --ssid "MyWiFi" \
  --password "SuperSecret" \
  --target-ip 192.168.1.42 \
  --target-port 5005 \
  --node-id 0 \
  --edge-tier 2 \
  --pres-thresh 0.6 \
  --fall-thresh 2.0 \
  --vital-window 64 \
  --vital-interval 200 \
  --subk-count 16

Key parameters include:

• --edge-tier 2: Enables full AI processing on the device (tier 0 = raw only)
• --pres-thresh and --fall-thresh: Presence and fall detection thresholds used by edge security modules
• --target-ip and --target-port: The server address (default UDP port 5005)

The provisioner generates a CSV NVS map and calls nvs_partition_gen.py to create the binary blob, then flashes it to the device (implementation in provision.py lines 33-71 for CSV creation, 74-104 for binary generation, and 118-135 for flashing).

Running the RuView Multistatic Server

The Rust server ingests UDP CSI frames from all nodes, performs multi-band fusion, multistatic viewpoint fusion, coherence gating, and Kalman tracking.

Launch the server:

cd rust-port/wifi-densepose-rs
cargo run -p wifi-densepose-sensing-server --release \
  --listen-udp 0.0.0.0:5005 \
  --listen-http 0.0.0.0:3000

The server pipeline includes:

• Multi-Band Fusion (wifi-densepose-signal/src/ruvsense/multiband.rs): Merges per-channel CSI into MultiBandCsiFrame (lines 39-55)
• Phase Alignment (wifi-densepose-signal/src/ruvsense/phase_align.rs): Removes LO phase offsets using a Neumann solver
• Multistatic Fusion (wifi-densepose-signal/src/ruvsense/multistatic.rs): Cross-node attention with ruvector-attn-mincut (ADR-029-p4)
• Coherence Gating (wifi-densepose-signal/src/ruvsense/coherence.rs and coherence_gate.rs): Computes Z-score coherence and decides Accept/PredictOnly/Reject/Recalibrate (enum GateDecision lines 58-67)
• Pose Tracking (wifi-densepose-signal/src/ruvsense/pose_tracker.rs): 17-keypoint Kalman filter with Re-ID (cost function lines 94-98)

Verifying and Visualizing Results

Connect to http://localhost:3000 to access the Observatory UI, which displays:

• 3-D skeleton overlay with 17 keypoints
• Real-time vital-sign plots (breathing rate, heart rate)
• Presence heat-maps and persistent field fingerprints

The UI consumes WebSocket data from /ws. For custom integrations, query the REST API at /api/v1/pose, which returns JSON matching the internal Pose struct defined in pose_tracker.rs (lines 77-84).

Summary

• Deploy 4-6 ESP32-S3 nodes around the room perimeter to create a multistatic mesh with 360-degree coverage.
• Flash the firmware from firmware/esp32-csi-node/main/csi_collector.c using esptool.py or the ./install.sh script.
• Provision each node with scripts/provision.py to set unique node IDs, Wi-Fi credentials, and sensing thresholds.
• Run the Rust server (wifi-densepose-sensing-server) to fuse multi-band CSI, perform multistatic fusion, and track 17-keypoint poses.
• Visualize results via the Observatory UI at port 3000 or consume the REST/WebSocket API for custom applications.

Frequently Asked Questions

How many ESP32 nodes are required for 360-degree coverage?

You need at least 4 nodes for baseline coverage, placed at wall midpoints approximately 2 meters apart. This configuration generates roughly 12 unique TX-RX links. For production deployments requiring higher SNR and multi-person separation, use 6 nodes to create 30 redundant links (N·(N-1)), as specified in ADR-029.

What is the difference between edge-tier 0 and edge-tier 2?

Edge-tier 0 streams raw CSI frames without on-device processing, requiring the server to handle all computation. Edge-tier 2 enables full AI processing on the ESP32, including presence detection, fall detection, and vital-sign extraction, reducing bandwidth and server load. Configure this via the --edge-tier flag in scripts/provision.py.

How does the server handle synchronization between multiple nodes?

The server implements Time Division Multiplexing (TDM) scheduling defined in rust-port/wifi-densepose-rs/crates/wifi-densepose-hardware/src/esp32/tdm.rs. Each node transmits null-data-packets in assigned slots while others listen, creating deterministic TX/RX pairs. The firmware in csi_collector.c handles the slot timing, while the server performs multistatic fusion using ruvector-attn-mincut attention mechanisms to align and fuse data from all viewpoints.

Can I use 5 GHz Wi-Fi instead of 2.4 GHz?

Yes. While the default configuration uses 2.4 GHz channels 1, 6, and 11 for channel hopping (configured in csi_collector.c lines 66-73), the firmware supports 5 GHz operation. You must modify the channel list in csi_collector.c and ensure your ESP32-S3 variant supports the 5 GHz band (note: standard ESP32-S3 is 2.4 GHz only; you may need ESP32-C6 or ESP32-C5 for 5 GHz). The server handles multi-band fusion automatically via the MultiBandCsiFrame structure in multiband.rs.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}