Deploy the Network

Move the trained weights from your laptop to the robot.

You can copy the network weights to the robot by executing the following command from the saved directory on your laptop:

scp steerNet.pt pi@192.168.50.5:~/RVSS2019-WS/on_robot/deploy/

Update the deploy file.

On the robot, in directory ~/RVSS2019-WS/on_robot/deploy/ there are two files to help you with deploying the network on the robot.

deploy_folder.py is an example of how to load images from a folder and feed them to the network. You may find this useful for debugging.

deploy0.py is an example of how to load an image from the camera, feed it through the network, and use the result to control the robot's driving angle. This is an example of a regression based network, so if your network was trained for classification, make sure to update this file!

Test the robot on a track.

Now you're ready to test the robot on some road configurations! Remember, you're being assessed on how quickly and accurately the robot can follow the road. If you're not happy with your robot's performance, try to figure out why. Is the controller for different steering commands the problem? Or is it the output from the network - do you need more data for training?