Robo-Drive

• 

  lane detection from a real world sample

• 

  lane detection

• 

  lane detection

• 

  the undead corpse of the rc the computer would control

• 

  a raspberry pi pico, which was meant to be a method to view the lanes, and run a portion of the computation

• 

  the raspberry pi that was meant to be the brains of the operation

Inspiration## Inspiration

After having won the third prize in HackedBeta by building a smart traffic controller, we were enthusiastic to pursue our attempts to solve real-life problems with technological innovations. We decided to build a car that would identify and abide by posted traffic rules and be able to travel within a lane.

What it does

This self-driving car uses a machine-learned algorithm to identify posted traffic signs and drive by abiding by the traffic rules. This car also has lane-detection software that ensures that self driving within a particular lane.

How we built it

We built this using a machine-learning algorithm with Tensor flow for identifying posted traffic signs. We a raspberry pi chip as a controller. We used a RC car, and modified the RC remote chip to be directly controlled by the Ras

Challenges we ran into

We were only able to procure a fisheye lens module for the raspberry pi controller, so we had to create a program to correct the distorted image from the fish eye lens for the Machine learning algorithm to be able to identify posted traffic signs. Insufficient processing power to train the model on a larger dataset, along with hardware limitations, which made it difficult to train the model, along with forcing us to compromise on how we would let the machine "see

Accomplishments that we're proud of"

Being able to identify and overcome specific problems that we encountered, all while this being our first times dealing with machine learning, and computer vision.

What we learned

how to train machine learning models, have the computer sense what it would encounter, how to work efficiently on a time crunch, teambuilding skills, and above all, how to have fun in a high pressure high stress environment.

What's next for Robo Drive

Hopefully implementing a proper method for computer vision, and training the machine learning model on a larger dataset, for longer periods of time, resulting in higher accuracy. once this is finished, we'd move on to upgrading the hardware and conducting more realistic tests for efficacy and reliability.

Built With

• machine-learning
• opencv
• python
• raspberry-pi
• tensorflow