The Covid-19 pandemic has highlighted the need for home monitoring and support systems for the vulnerable sections of our population (e.g. senior citizens). A 3D printed home surveillance robot can monitor and automatically broadcast emergency alerts regarding urgent care needs of senior citizens (e.g. fall detection, intrusion detection etc). An important reason to 3D print a robot was the exceedingly high cost of home-monitoring-robots in the marketplace (ranging from $3000 to over $10,000). Thus, with all the tools and knowledge gained from the computational fabrication course at UCSB, I was motivated to bring my skills to life with my own affordable robot.
Capabilities
Detect objects and people
Oni houses one USB camera which captures the video feed of the world around it. It also categorizes and performs inference on what it sees on-the-edge using SSD-net v1 model built on the COCO dataset.
Perform inference, capture images and send alerts via email
Oni detects if someone in the house has fallen down, and sends a real-time email alert with an attached image (depicting the emergency) to a person’s emergency contacts.
Video Streaming
Oni uses its camera to infer and broadcast live video streams to local devices connected over a common network.
Extensibility
All the major components of Oni are designed to be integrated with screws. This makes Oni’s body modular and vertically extendable in case the space requirement increases to incorporate more hardware and sensors to expand Oni's capabilities.
Movement
A motionless robot is very restricting. Hence, I gave Oni motion support with continuous tracks that are not only aesthetically appealing but also functionally robust. Compared to wheels, continuous tracks with treads have an optimized traction system which enhances power delivery and efficiency for Oni's motors.
TPU printed tracks
Oni has TPU printed tracks which are more or less similar to rubber material. As a result they have reduced PSI on the ground as compared to PLA printed tracks and hence, reduces impact on the ground and creates less noise while movement.
Battery life
Oni runs on a 12V battery with 5200 mAh capacity. It supports four embedded boards, a USB WiFi adapter, a USB camera, three servo motors and two 12V DC motors. Oni can run continuously for 1.5 to 2 hours which can be extended with a higher capacity battery.
Features
Oni is packed with 4 embedded boards that control the servo and DC motors as well as on-the-edge visual computing and processing. It reduces latency (which would have happened if connected to cloud in contrast) and protects the user's privacy as the data is processed by One locally (and not sent to the cloud).
Processor
NVIDIA Jetson Nano with 4GB RAM
Computer Vision
- 180 degrees of horizontal head movement
- Logitech USB Camera
- Jetson-inference algorithm combined with logic to detect fall
Movement
Track-driving tread system
Connectivity
2.4Ghz WiFi Connection
Design
- Expandable with screw system
- Vents to release heat exerted from embedded boards housed inside
- Comprehensive head design that can encase both USB camera as well as Raspberry pi camera and a speaker as future scope
Build and Dimensions
Project’s Benefit and Future Scope
The major benefit of this project was having a home surveillance robot with comparable features of personal robots available in the market at a much lower price. Furthermore the modular nature of 3d printing and assembly leaves room for more capability and feature expansion in the future. Oni currently meets its desired objectives, i.e. it recognizes fall in real-time and sends out emergency alerts with images over email. I’ve gained major hands-on experience in different types of fabrication involving 3D printing with different materials such as PLA and TPU, soldering, creating joints and integrating 3D printed parts with motors and embedded boards. The scope of this project is huge. To list down a few points, I am looking forward to extend Oni to include voice command support (e.g. “Oni, come here!”, “Oni, get me help”), lidar and ultrasonic sensors for depth mapping and autonomous navigation around the home and a lightweight camera for a faster and more flexible neck movement.