Driver Software Suite

Driver Software Suite — How to Build Your Own Autonomous Vehicle

From the very beginning, when we started working with autonomous technologies, we decided to build a modular system capable of automating any wheeled vehicle. As is often the case, universality comes at the cost of additional development time. However, it allows us to avoid being tied to a specific form factor. We simply adapt the motion kinematics to different types of wheeled platforms.
For example, a cleaning robot uses a tank-style drive system, a stacker has a single 360-degree drive wheel, and a car uses front differential drive, which is much more complex mathematically, but still achievable.

Kinematics is the first stage of configuration. Next comes integration with the drive system. In robots, this can be either a synchronous or asynchronous motor. Each motor has its own controller, and it is crucial that it includes a PID controller capable of maintaining the desired speed under varying loads.

With synchronous motors, we use the VESC controller — a very convenient solution compatible with all motors of this type. As for asynchronous motors, it’s more complicated: there is no universal controller. Existing solutions, such as Curtis, are closed-source, so you often have to dive into the details — probing each pin to find the correct one for signal transmission.

Next comes sensors and inertial measurement units. While working with an accelerometer is straightforward — you simply receive data and process it — each LiDAR has its own documentation and integration approach. Libraries for one LiDAR won’t work with another. In addition, there are 3D cameras, parking sensors, solid-state LiDARs, liquid level sensors, actuators for lifting brushes, and many other auxiliary components.