FIDO | Devpost

Kristin Farris CS '20, Easton Bornemeier CS '19, Stephen Agee EE '19, Tanner Lucas EE '19
FIDO

Inspiration

We created FIDO for a fun home-robotics application of computer vision. Additionally, the methods used on FIDO have various applications in larger areas related to search and rescue, such as in aerial drones.

What it does

FIDO is a robot dog that is designed to play fetch with you. It has the ability to track a tennis ball and it's owner for a fetch and retrieval game as well as keep-away.

How we built it

The robot chassis was built using a modified Elegoo robotics kit and legos. The robot uses a mounted camera that sends data to an on-board raspberry pi running OpenCV. Using color thresholding to identify the Tennis Ball, the pi creates a binary image, which is then used to track the distance and direction of the ball from FIDO. This information is sent to the Arduino, which uses feedback control to move the motors and correct the error. Once the tennis ball is acquired, FIDO returns to the owner, who is marked with a neon green shirt.

Challenges we ran into

Being able to identify when the ball was picked up and when it had been returned to the owner was a tricky task. We ended up using contour detection on the edges of the camera frame to check when the ball/person had reached viewable extrema.

Accomplishments that we're proud of

We are very proud of the ball-tracking aspect of the computer vision. FIDO can see the tennis ball at a max range of almost 50 ft and track to it without fail.

What we learned

We learned that making something that is geared towards entertainment is entertaining in itself!

What's next for FIDO

Although the current prototype only allows for fetch using a tennis ball, the same principles can be applied to larger robots, and even unmanned aerial drones. Future applications extend beyond puppy playtime to include search and rescue and hazmat extraction using computer vision. This would allow FIDO to be both entertaining and useful in creating a better, safer world. FIDO - the future of man’s best friend.

Built With

Submitted to

XBuild 2018

Created by

I worked on the computer vision color detection and contour processing, as well as the I2c data communication. I also worked with the motor shield interface and programatically controlled the motors using the ball data.

Easton Bornemeier
I worked on implementing color threshholding for FIDO’s computer vision following program. Additionally, I constructed the pickup mechanism for our tennis ball retrieval method.

Kristin Farris
Steve and I worked on setting up the raspberry pi with an instance of Arch Linux. This included installing updates, accessing the pi through ssh, connecting the pi to the local network, and setting its hostname and security information. We also pair programmed the Arduino motor driver and assembled the chassis for the robot.

tlucas96
I worked with Tanner to assemble the robot enclosure. After assembly, we wrote the basic driver code for the motors on the Arduino to allow the robot to move around. Once this was done and I2C was set up, I wrote startup services in Arch Linux on the PI to allow the robot start running on boot. Finally, I edited the video to highlight the features of our bot.

Stephen Agee