discussion / Open Source Solutions  / 15 June 2025

SwarmGuard: low-cost autonomous mesh system to detect poaching threats and protect wildlife


Hello WILDLABS Community!

I’d like to share an open-source, non-commercial concept that could be useful for conservation teams, rangers, or students working with wildlife monitoring in remote areas:

---

##  What is SwarmGuard?

SwarmGuard is a **modular, ultra-low-cost, solar-powered mesh of autonomous sensor nodes** designed to detect poaching threats like:

- Gunshots  
- Predator roars  
- Human speech or presence  

Each node is small, quiet, and smart — like a tiny forest sentinel that wakes up when something dangerous happens.

---

##  Core Components (per node):

- **ESP32-CAM**: microcontroller + camera (or ESP32-WROOM for cheaper audio-only nodes)  
- **MAX9814**: high-gain microphone with automatic level control  
- **SX1278 LoRa module**: for long-range, low-power communication  
- **TP4056**: battery charge controller  
- **18650 Li-ion battery**  
- **5V solar panel (~120–200 mA)**  
- **Waterproof plastic enclosure (IP65)**  

 Total cost per node:  
- With camera: ~$35–40 USD  
- Without camera (basic audio relay): ~$20–25 USD

---

##  How it works:

- Node stays in **deep sleep** (uses <1 mA)
- **Wakes up on sound threshold**
- TinyML model classifies sound (gunshot, roar, voice, wind, etc.)
- If confirmed → takes a photo (optional), stores event, and sends **LoRa alert**
- Signal hops from node to node until it reaches:
  - base station (with LoRa → WiFi bridge),
  - relay drone,
  - or GSM gateway

Nodes can also periodically transmit:
- Battery level
- Noise levels
- Presence history
- Coordinates (if GPS module added)

---

##  Smart Ideas:

- Combine **"cheap relay nodes"** (no camera) with **"photo nodes"** in strategic spots  
- Drop cheap nodes from drones or throw into bushes — they form a mesh  
- Central server can **analyze trends**, filter false positives, re-train models  
- Whole network is **resilient** — if one node fails, others keep working

---

##  Use Cases:

- Forests with limited human access  
- Wildlife corridors or buffer zones  
- Parks with limited ranger staff  
- Cross-border patrol zones  
- Anti-poaching operations for rhinos, tigers, elephants, etc.

---

## Pros:

-  Very low cost — mass deployable  
-  Fully solar-powered, autonomous  
-  Smart sound detection (gunshot, roar, speech)  
-  Mesh networking via LoRa (no internet needed)  
-  Optional camera capture  
-  Open-source hardware + firmware

---

##  Limitations:

- LoRa range can vary (trees, hills affect it)  
- Cheap plastic case may degrade without weatherproofing  
- TinyML model needs fine-tuning for specific regions  
- No GPS by default (can be added if needed)

---

## ❤️ Why share this?

This is not a commercial product — just an open idea.  
If someone out there can save animals, track threats, or scale this up — I’m happy to contribute, collaborate, or hand over everything.

If you want help with:
- PCB layouts
- Arduino/PlatformIO firmware
- Edge Impulse training setup
- Wiring diagrams

…I’ll be happy to assist.

Let’s build a quiet, invisible mesh that watches over the wild — and helps those who can’t ask for help themselves. 🐘🕊🌱

If you find this idea useful, interesting, or if you end up building anything based on SwarmGuard — I’d truly love to hear about it.

Please feel free to message me here, or email me at:
📩 [email protected]

Let’s protect wildlife together.

– Aleksey (Russia)
 

SwarmGuard_Project_Overview.pdf


Good morning Aleksey,

Your idea is very similar to an idea I had -- I guess we are experiencing some convergent evolution.   I'd be very interested in understanding your project more.   I'd be curious to know if you've done much testing with Lora and dense foliage.  I'm just curious to know if you've hit any real limitations.  Reading your description, it sounds like perhaps you are using some mesh networking.   Anyway -- it's nice to meet you and see we had similar ideas.

Chris

Good concept and the component choices are sensible for a low-cost prototype. A few technical observations from working with similar architectures in remote African deployments:

On the ESP32 deep sleep current — the ESP32-WROOM in deep sleep draws ~10-20μA in practice, which is fine for battery life calculations, but the TP4056 charging IC has a quiescent current of ~55μA in standby. This means the charging circuit itself can dominate your sleep current if the solar panel voltage is too low to trigger charging. Worth profiling the actual system current rather than relying on datasheet deep sleep figures alone.

On the MAX9814 — it's a good low-cost choice but has a relatively high quiescent current (~3mA continuously) for a mic preamp. If you're using sound threshold waking via an interrupt, the mic circuit needs to stay powered during sleep, which significantly impacts your power budget. The ICS-43432 I2S MEMS mic combined with a hardware comparator for wake-up threshold detection is a more power-efficient alternative for this use case.

On the SX1278 LoRa mesh — the SX1278 supports raw LoRa but not LoRaWAN natively. For a mesh topology you need a protocol layer on top — Meshtastic or a custom implementation. Worth being explicit about this in the architecture, as it affects how you integrate with existing ranger communication infrastructure.

On the IP65 enclosure in African savannah — IP65 protects against dust and water jets but not against the daily thermal cycling that drives condensation inside sealed enclosures. A Gore-Tex vent membrane port is worth adding to any permanent outdoor installation.

Interesting project overall — happy to discuss the power architecture in more detail.