At the Technical University of Berlin (TUBerlin), nearly eighty students form the Formula Student Team, or FaSTTUBe. Each year, they construct racing cars to compete in international design competitions. Since 2019, the team has focused on developing an autonomous driving system that enables their car to navigate racetracks completely independently.

The challenge lies in the racetrack’s tight corners and chicanes, demanding a vision system with low latency and high image quality for reliable visual navigation. Last season marked a significant breakthrough, largely thanks to Allied Vision Alvium CSI-2 cameras.

Low Latency Image Transfer

In autonomous mode, the car identifies drivable routes using cone markers placed at intervals along track edges. Precise cone localization requires capturing images with minimal distortion and transmitting them quickly to the NVIDIA Jetson AGX Xavier’s graphics card via a low-latency connection. There, neural networks preprocess these images efficiently for spatial analysis.

However, an initial prototype using USB cameras proved insufficiently fast for the race car’s dynamic environment.

High Quality Images

The FaSTTUBe team selected two Alvium 1800 C-319c cameras as a critical component of their autonomous pipeline. These cameras feature a global shutter sensor and MIPI CSI-2 interface, which prevents rolling shutter effects at high speeds by eliminating distortion.

Moreover, the compact form factor proved advantageous for integration into their enclosed waterproof sensor box.

Higher Precision

Thanks to Allied Vision’s cameras, FaSTTUBe implemented an advanced computer vision concept that achieved two major improvements: a 150% increase in cone detection frequency and a remarkable 400% extension of the detection range. These enhancements significantly boosted performance during autonomous driving phases.

Small, High-Performance Industrial CSI-2 Camera

The Alvium 1800 C-319 with Sony IMX265 sensor delivers an impressive 54 frames per second at 3.2 MP resolution. It utilizes the powerful MIPI CSI-2 interface (with up to four lanes) for high bandwidth.

Onboard ISP capabilities support advanced image processing, freeing up embedded computing resources for other tasks. Its compact design and intelligent power management allow seamless integration into low-power or battery-powered systems.

  • Easy integration into enclosed waterproof sensor boxes
  • Elimination of distortion through global shutter technology
  • 150% increase in cone detection frequency
  • 400% extension of detection range

Last Updated: 2025-09-04 16:19:34