Skip to content
Permalink
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
78 lines (67 sloc) 6.26 KB

BUILD YOUR OWN SELF-DRIVING CAR

The goal of this repository is to present the topic of retrofitting a hobbyist car for the purpose of driving it autonomously, as well as to collect a number of resources for making that goal a reality. We hope to see the rise of the Homebrew Computing Club for the autonomous vehicle industry, an opportunity for amateurs, researchers, and DIYers to participate in this burgeoning field in a way that is accessible, affordable, and fun.

MOOCs have given us unprecedented access to classroom learning, and from this democratization of knowledge, we’ve already seen a huge surge in creativity, collaboration and experimentation that comes from an open-source culture. We want to extend that all the way to the hardware and the sensors and prove that the software we write and the models we train can really hold their own under real world conditions.

WHY?

  • Democratize the technology
    • The architecture of a self-driving RC car is not, in principle, all that different from that of a full-sized vehicle. This puts the full autonomous vehicle architecture, from the sensors, micro-controllers, motors, etc. to the localization, planning and control algorithms, within reach of anyone who wants to learn.
    • Online courses have given us unprecedented access to classroom learning, and we’re already seeing a huge surge in creativity, collaboration, experimentation and innovation that comes from a more open-source-centric culture. Let's extend that all the way down to the hardware. There’s no reason you shouldn’t be able to implement what you learned in class yesterday on your own self-driving car today, and then share what you learned in the process in a blog post tomorrow!
  • Real world data beats simulation
    • If you have taken or are considering taking any of the available courses from Udacity, Coursera, or elsewhere, you will have the opportunity to learn about and write the algorithms for things like steering control, obstacle detection, and path planning. And in most cases, you'll run this code against canned data or a simulator. But having access to a physical platform allows you to move from code that works under idealized conditions to code that functions in the real world with all its uncertainty and performance constraints. Not only is this more of a challenge, having this experience makes you valuable.
  • Cost
    • On a cheap, small scale platform, you have the freedom to explore and experiment in ways that researchers working on full-sized cars cannot.
  • Safety
    • Fear of catastrophic failure leads us to optimize for safety and for preserving precious equipment without really pushing the boundaries of what the technology can do.
    • Crashing an RC car is not a catastrophic event in the way that crashing a full-size car is.
  • Push the boundaries of Sensing, Perception and Control
    • There’s a quote by Mario Andretti, “If everything seems under control, you’re not going fast enough.” It's only when we push the boundaries of our technology that we begin to understand where our assumptions break down.
    • We want to push the limits of what our software and hardware can do. We want to move fast, break things, learn, and have fun!

WHAT DOES IT TAKE TO BUILD?

  • Chassis
  • Sensors
    • Camera: visual perception
    • Lidar: spatial/depth perception with pulsed laser range sensors
    • IR Depth Imaging: spatial/depth perception using projected infrared light and camera
    • IMU (Inertial Measurement Units): measures and velocity and angular rate using a combination of accelerometers and gyroscopes
    • Radar, Sonar, etc.
  • Controller
    • The sensors are interfaced with a computer or computers, onboard and remotely via Wifi, which gather than sensor data and relay it to the software
      • Raspberry Pi (Zero or 3) - Cheap, but less powerful, often delegates perception and planning to another computer, such as a laptop
      • Nvidia Jetson Series - More powerful micro-controller for onboard computation
      • Android, Beaglebone, etc.
  • Software
    • Open source software like ROS, OpenCV, Tensorflow, etc.
    • Perception and planning algorithms process and combine the sensor data and send back steering and throttle commands to the car

HOW DO I GET STARTED?

Fortunately, there are already a number of guides we can follow to get started.

OTHER RESOURCES

COMPETITIONS

Competitions are a great way to motivate the problem, focus our efforts, and get exposed to what people are doing in the world.

You can’t perform that action at this time.