Tarsier Prototype vs. Competitors: What Sets It Apart

Building the Tarsier Prototype — Features, Specs, and Roadmap

Overview

The Tarsier Prototype is a compact, low-power edge-computing device designed for sensing and local inference in constrained environments. It targets rapid prototyping for robotics, environmental monitoring, and smart sensors, prioritizing modularity, low latency, and battery efficiency.

Key features

• Modular I/O: Swappable sensor and communication modules (I2C, SPI, UART, GPIO).
• Edge ML acceleration: Onboard NPU for common tinyML models (classification, anomaly detection).
• Low-power modes: Deep sleep with wake-on-interrupt and scheduled wake timers.
• Rugged form factor: IP54-rated enclosure option with vibration damping.
• Secure boot & OTA: Signed firmware updates and hardware root-of-trust.
• Local storage: 4–32 GB eMMC or microSD options.
• Flexible power: LiPo, USB-C PD, and optional solar input with MPPT.
• Time-sync: RTC with optional GNSS/time-sync module.
• Developer-friendly: USB debugging, SWD, and open schematics/firmware.

Typical specifications (baseline prototype)

• Processor: 32-bit ARM Cortex-M7 @ 400 MHz
• NPU: 0.5 TOPS fixed-point accelerator
• RAM: 512 KB–2 MB SRAM (depending on variant)
• Flash: 2–8 MB onboard + microSD slot
• Wireless: Optional 802.15.4 / BLE 5.2 / Wi‑Fi 4 module
• Power: 3.7 V LiPo, USB-C PD input, 100 mA idle, <10 µA deep sleep
• I/O: 2× I2C, 2× SPI, 4× UART, 12× GPIO, ADC channels
• Sensors (optional): IMU, temperature, humidity, light, microphone
• Size: ~60 × 40 × 15 mm

Roadmap

1. Prototype v0 — Proof of concept
   • Validate core CPU + NPU functionality
   • Basic I/O and power management
   • Onboard bootloader
2. Prototype v1 — Feature-complete dev board
   • Modular connectors for sensors and comms
   • Improved power efficiency and RTC
   • Initial enclosure and mounting points
3. Pilot v1 — Field testing
   • Ruggedized case, IP54 testing
   • OTA update pipeline and security audit
   • Battery life optimization with real workloads
4. Production v1 — First production run
   • Finalized BOM, certification (FCC/CE)
   • Documentation, SDK, and sample models
5. Future iterations
   • Higher NPU throughput, additional comms (LoRa/5G)
   • Miniaturized variants and integration kits

Development timeline (example)

• Months 0–2: Design & schematics
• Months 3–5: PCB fabrication & bring-up
• Months 6–8: Firmware & NPU integration
• Months 9–11: Field pilot and iterations
• Month 12: Production release

Risks & mitigation

• Thermal limits: Use thermal vias and metal shielding; throttle heavy workloads.
• Power draw: Profile workloads; add dynamic voltage scaling.
• Security vulnerabilities: Hardware root-of-trust, code signing, regular audits.
• Supply chain: Multiple vendor components and last-time-buy planning.

Suggested SDK & tooling

• Cross-toolchain: arm-none-eabi GCC, CMake
• ML: TensorFlow Lite Micro, CMSIS-NN
• Debugging: OpenOCD + SWD, serial console
• CI: Automated unit tests, firmware signing pipeline

Quick start (developer flow)

1. Flash bootloader via SWD.
2. Load sample firmware with sensor drivers and tinyML model.
3. Attach sensor module and power via USB-C.
4. Run inference locally; profile power and latency.
5. Configure OTA and secure update testing.

Date: February 8, 2026