# Title Team Members TA Documents Sponsor
62 Autonomous Pothole Detection and Cataloging for Bikes
Andy Sun
Harshvardhan Bhatia
Jesse Chen
Xinrui Zhu appendix
Title: Autonomous Pothole Detection and Cataloging for Bikers

Link to original idea thread:

Potholes are an issue which plague cities all over the country and over the world. While damaging to cars, potholes can also be particularly dangerous, even fatal, for bikers and can lead to millions of dollars in lawsuits for a city if not patched [1].

This product would attach onto a bike's handlebars and utilize both computer vision techniques and accelerometer/ultrasonic sensor data to detect potholes and automatically record their location (via GPS) to a database that can be used by cities to determine problem areas. In addition, there would a button that a biker could press to add a pothole if it isn't automatically detected. There would also be a local copy of the database on the device which would enable the device to alert the user if they were approaching an area with potholes reported. An armband that the user wears would issue both a haptic and hearable alert.

Current embedded computer vision techniques have been tested at around 70-80% accuracy in broad daylight conditions [2], therefore we cannot entirely rely upon them. However, it does provide a means of cataloging potholes that bikers do not run over and can even serve as a an early warning system if we are able to detect a pothole early enough. The accelerometer and ultrasonic sensors can serve to catalogue potholes at night and hard-to-see potholes during the day, and in general have much better accuracy. In addition, we can try to catalogue the potholes based upon their severity, which we would estimate based upon our sensor data. We may need to have two MCU's: one dedicated for image processing and the another handling everything else.

There are somewhat similar products out there for cars (no commercial products from what I can tell, only devices created for research),, but no such products for bikes and bike paths. A pothole can be much more devastating for a biker than for a vehicle driver, so we believe this is a good problem to focus on.


Amphibious Spherical Explorer

Kaiwen Chen, Junhao Su, Zhong Tan

Amphibious Spherical Explorer

Featured Project

The amphibious spherical explorer (ASE) is a spherical robot for home monitoring, outdoor adventure or hazardous environment surveillance. Due to the unique shape of the robot, ASE can travel across land, dessert, swamp or even water by itself, or be casted by other devices (e.g. slingshot) to the mission area. ASE has a motion-sensing system based on Inertial Measurement Unit (IMU) and rotary magnetic encoder, which allows the internal controller to adjust its speed and attitude properly. The well-designed control system makes the robot free of visible wobbliness when it is taking actions like acceleration, deceleration, turning and rest. ASE is also a platform for research on control system design. The parameters of the internal controller can be assigned by an external control panel in computer based on MATLAB Graphic User Interface (GUI) which communicates with the robot via a WiFi network generated by the robot. The response of the robot can be recorded and sent back to the control panel for further analysis. This project is completely open-sourced. People who are interested in the robot can continue this project for more interesting features, such as adding camera for real-time surveillance, or controller design based on machine learning.

Project Videos