Team Jetracer is testing and using the Jetracer framework on robotic race cars and the Donkeycar Simulator.
Fully Autonomous Test Run
Jetracer Autonomous DIY Optimized
The Triton AI Jetracer team has cracked the code for the NVIDIA Jetracer DIY, optimizing how the Jetracer steering and throttle decisions made with PyTorch. The Triton AI Jetracer DIY robot now has the fastest lap time at 11.32 seconds because of the optimized training algorithm with carefully selected training images. Watch and learn a few Jetracer techniques to implement on your own Jetracer! Ask any questions Jetracer questions about our Jetracer!
Fully Autonomous part 1
Fully Autonomous part 2
Triton AI Jetracer DIY
Waveshare Jetracer Pro
– Uses a multiplexer to relay the signal from the radio controller to steering servo and motor.
– When channel 3 is switched on the multiplexer switches to the PWM. This is how the Jetracer controls the throttle and steering.
– Uses standard Jetracer github repo and Adafruit-PureIO library for I2C with PWM.
– Uses Waveshare “expansion board” that is a custom multipurpose PCB that distributes power to the Jetson and motors from 8.4v batteries.
– The expansion board also acts as the PWM interfacing the jetson to the motors via singal.
– Expansion board also uses OLED to display IP address when connecting via Jupyter Notebook.
– Uses Waveshare/Jetracer github repo fork and UPDATED Adafruit-PureIO library.
– Need to create Waveshare repo folder for more library resources for the expansion board.