Donkey Car

Complete the parts list for the selected car

The hardware guide lists the car, Pi, camera, storage, USB power bank, PWM board and fasteners, but gives printed mounting parts an unpriced placeholder. It also discusses chassis-specific adapters hosted elsewhere. Those omissions prevent the table from establishing a complete delivered budget. This review has not acquired and reconciled all external print files or their licenses. [S005]

The guide’s hardware table names a Pi 5B with 8GB RAM and a 64GB card, while the software page recommends a Pi 4 or 5 with 4GB RAM and a 128GB card. Treat these as different documented choices to resolve, not as one exact bill of materials. The RC vehicle’s own steering servo, ESC and motor remain purchased components. [S005] [S006] [S007]

The saved installation guide contains a version conflict

The Pi setup page opens by saying Donkey Car 5.4 requires 64-bit Raspberry Pi OS Trixie, but its installation heading and following sentence still name Bookworm. Its environment command uses Python 3.13. The software release metadata independently requires Python at least 3.12 and below 3.14, with separate Pi and host-training dependency groups. [S007] [S009]

The guide’s developer installation also checks out main, which does not reproduce the pinned release automatically. Select and record the OS image, Python environment and software revision together before treating the instructions as a repeatable setup. No installation or compatibility test was performed here. The guide also says TensorFlow Lite models need regeneration across the versions it discusses; existing model files should not be assumed compatible. [S007]

Calibrate the actual steering and ESC

The calibration guide directs builders to raise the car off the ground, find steering endpoints without forcing the servo, and save the values in myconfig.py. The ESC needs its own forward, neutral and reverse settings; the guide describes a reverse-neutral-reverse sequence for RC controllers that require it. Its example PWM numbers are starting points, not measurements of your car. [S006]

After initial calibration, the guide checks straight driving and adjusts the steering settings. Preserve those configuration values with the chosen drivetrain and PWM channel mapping. A working camera and manual driving configuration are prerequisites to collecting useful demonstrations, not evidence that an autonomous policy will generalize to a new track. [S006] [S007] [S008]

Training and hardware openness are separate questions

The documented learned-driving route uses data collected on the car and a separate host computer for training. The inspected package has distinct Pi inference and host training dependencies. This profile does not select or validate a trained model, dataset or autonomous-driving result. [S008] [S009]

The software repository carries an MIT license. That license does not establish rights to the separately hosted build documentation, external printed parts or purchased RC components. The documentation archive was preserved, but no license file was identified in its inventory. No assembly, calibration, model training or driving test was performed for this entry. [S003] [S010]