This week I focused on background research to validate the feasibility of the proposed sensing, computation, and communication approach for the Shade.ai system. I reviewed the computer vision software stack to confirm that an OpenCV-based pipeline using lightweight detection models with higher-rate tracking can meet our response-time and accuracy requirements for single-user following. I also researched hardware requirements for CV, including camera modules, CSI connectors, and mounting considerations, confirming that CSI-based cameras are the most suitable option due to their low latency and power efficiency. Additionally, I explored app-level communication options and compared Wi-Fi and Bluetooth for user control, noting that Wi-Fi is better suited for real-time control and status updates.
In parallel, I investigated drone hardware considerations and obstacle avoidance requirements. This included reviewing payload, power, and endurance constraints that support a lightweight, low-power companion computer approach. I also compared 2D versus 3D obstacle sensing and determined that simple ToF-based obstacle detection is sufficient for our low-speed, low-obstacle operating environment, avoiding the complexity and cost of full 3D mapping. Overall, this research supports a Raspberry Pi–based MVP design with lightweight sensors that balances performance, safety, and system simplicity.