Inspiration
Everyday inconveniences can waste time and diminish life’s enjoyment. Something as simple as dropping your phone under the couch and spending 10 minutes awkwardly retrieving it isn't exactly a fun pastime. This inspired us to create a solution for recovering objects from hard-to-reach places.
What It Does
Our robot uses a camera to detect objects based on RGB concentration and moves towards them while ensuring it doesn't overshoot, thanks to the camera and the ultrasonic sensor. Once positioned correctly, the claw mechanism lowers and picks up the object.
How We Built It
We assembled the robot using a kit, similar to adult LEGO, making the process structured yet flexible. The software was developed on a Raspberry Pi 4 using Python and relevant libraries. A key component was YOLO (You Only Look Once), a real-time object detection system, which enabled our robot to recognize and track objects for precise retrieval.
What We Learned
Setting up and configuring a Raspberry Pi, including SSH connections and server-based communication. As well as implementing Python libraries to enhance functionality and efficiency in real-world applications. Finally, Overcoming hardware and software challenges to build a functional, autonomous retrieval system.
Challenges We Faced
Firstly we had navigation issues with infrared sensors. Initially, we attempted to guide the robot using an infrared sensor to follow a black tape path, but this approach failed due to inconsistencies in surface reflectivity and sensor accuracy.
Additionally we had problems with object detection complexity. We explored a more advanced object detection system capable of recognizing multiple items by masking over them instead of just using RGB or bounding boxes. However, this method was impractical due to slow image processing and our limited experience in AI.
Accomplishments We're Proud Of
We successfully used the camera to dynamically update the robot’s position relative to the target object by processing real-time RGB data. Getting it to consistently track and retrieve a red ball—our test object—was a rewarding moment, as it demonstrated our ability to apply object detection technology in a new way.
Log in or sign up for Devpost to join the conversation.