What to work on for the June 17th DIY Robocars Race

Here are some goals to improve the Donkey V2 car. Completing these by the July 17th DIY Robocars Race will give us competitive race times and would prepare us begin doing object detection and avoidance.

1. Train an autopilot for the DIY Robocars track that use throttle. Everything exists to make this happen but better data is needed. Manual annotation of past datasets to speed up on the straightaways could work.
2. Train an autopilot that have a memory of a few frames.  This will help the car avoid the rouge command to turn the opposite direction for a single frame that gets if off course. Memory can be implemented by using a recurrent layer in the network. Implementation may require changing the training/driving code to provide sequences instead of individual frames.
3. Build at least one car with an odometer and document how to install it. * Currently the autopilot output is throttle. This is a problem because the speed of the car for any given throttle can vary depending on the angle of the road or how the battery state. This will require adding an additional variable to each frame. 
4. Build at least one car with LIDAR and document how to build it. * Obstacle avoidance is one of the basic requirements of self driving cars. LIDAR can be used see how close objects are to the car.  With LIDAR on at least one car, we'll learn what depth information can be captured on the track. 

* 3 & 4: The goal is to capture enough data from the cars with expensive/difficult sensors so that a neural net or visual heuristic can be trained to predict the output of these sensors using just the camera. This will enable other cars to install software versions of these sensors on their existing cars without the hardware. 

There are also improvements to the web interface that would help speed up training and testing. Currently there is a lot of command line steps people have to remember to build a working model. 

1. Button on the drive screen to start a new session.
2. Some way to rename existing sessions. 
3. Link download existing sessions in a zipped file or mp4 video.
4. Show summary of pilots when pilot is clicked. 
5. Link to download pilots in hdf5 format. 
6. Page to select sessions and train autopilots. 

If you'd like to contribute, please join our slack channel, create a github branch, make your improvements, and submit a pull request. These will goals will be made into issues on github as well.