The GAP9Shield is a state-of-the-art hardware module designed to enhance the performance of nano-drones, enabling advanced vision-based and ranging applications. Powered by the GAP9 System-on-Chip (SoC), this platform delivers unparalleled AI processing capabilities and energy efficiency. With its compact design and cutting-edge features, GAP9Shield revolutionizes the potential of nano-drones, particularly for indoor environments and close-proximity interactions with humans.
- Architecture: 9-core RISC-V cluster with AI acceleration (NE16).
- Performance: 15.6 GOPs for DSP, 32.2 GMACs for ML tasks.
- Energy Efficiency: 330 µW/GOP.
- Dynamic Scaling: Frequency up to 370 MHz, sleep power as low as 45 µW.
- Memory:
- 1.6 MB L2 RAM and 2 MB embedded non-volatile memory.
- Additional 256 Mbit PSRAM and 512 Mbit Flash.
- Resolution: 5 MP, supporting QSXGA to QVGA formats.
- Frame Rates: Up to 120 fps (QVGA).
- Interface: CSI2 single-line connectivity.
- Processor: ESP32 dual-core MCU.
- Wireless Capability: 2.4 GHz Wi-Fi and Bluetooth.
- Throughput: Up to 12 Mbps (UDP) with low power consumption.
- Directionality: Covers right, left, front, back, and upward directions.
- Range: Up to 400 cm with a sampling rate of 60 Hz.
- Dimensions: 50 mm x 27 mm (6-layer PCB).
- Weight: Approx. 6 grams.
- 20% higher frame rate for RGB images compared to conventional AI-decks.
- Optimized for AI workloads like YOLO object detection, MCL localization, and SLAM.
- Operates below 100 mW for most tasks.
- Designed for extended flight durations with lightweight, power-efficient components.
- Compatible with high-definition cameras via MIPI CSI2.
- Integrated Wi-Fi and BLE modules for seamless data streaming and communication.
- 15% lighter and 30% smaller than comparable setups.
The GAP9Shield is ideal for:
- Obstacle avoidance in confined spaces.
- Real-time object detection and mapping.
- Greenhouse farming and industrial inspections.
- Precise image recognition and navigation in cluttered environments.
-
Hardware: Contains the hardware files for the shield. 1.1 gap9_deck_v2: Contains the design files (Altium) for the shield. 1.2 ToF-FrontBack: Contains the design files (Altium) for the ToF PCBs (front and back). These need to be mounted first and then soldered to the main PCB. 1.3 ToF-LeftRight: Contains the design files (Altium) for the ToF PCBs (Right and left). These need to be mounted first and then soldered to the main PCB.
-
Firmware: will contain firmware files (future release).
-
Documents: Contains the repository documentation. 3.1 Schematics: Main PCB and accesory PCB schematics in PDF format. 3.2 Pictures: Images/3D models of the shield. 3.3 Diagrams: GAP9shield diagrams.
If you use the GAP9shield in an academic or industrial context, please cite the following publication:
@InProceedings{muller2024gap9shield,
author="M{\"u}ller, Hanna
and Kartsch, Victor
and Benini, Luca",
title="GAP9Shield: A 150GOPS AI-Capable Ultra-low Power Module for Vision and Ranging Applications on Nano-drones",
booktitle="European Robotics Forum 2024",
year="2024",
publisher="Springer Nature Switzerland",
address="Cham",
pages="292--297",
isbn="978-3-031-76424-0"
}
This project is open source.
The files within the following folder (and subfolders) are released under Solderpad v0.51 (SHL-0.51) (see LICENSE):
Hardware/
The files within the following folder (and subfolders) are released under Creative Commons Attribution 4.0 International
License (CC-BY-4.0) (see Documents/LICENSE):
Documents/