GigE Vision Driver Development

The Challenge

Daimler AG (Mercedes-Benz) and Robert Bosch GmbH required for their cooperation project Athena a high-speed image capture solution for their Advanced Driver Assistance Systems (ADAS) testing infrastructure. The system needed to simultaneously capture data from four GigE Vision cameras at 10 Gbit/s each, totaling 40 Gbit/s of raw video data. The existing solutions could not handle the massive data throughput required for real-time analysis in water-cooled Eurotech Linux systems deployed in test vehicles. The driver needed to implement the GigE Vision Control Protocol (GVCP) and GigE Vision Stream Protocol (GVSP) for four parallel camera streams without data loss.

Our Solution

We developed a custom Linux driver in C implementing the complete GigE Vision standard with support for parallel multi-camera operation. The driver features a discovery mechanism that recognizes all four cameras as individual devices over separate network interfaces, with each camera receiving its own GVSP stream handler. We implemented multi-threaded parallel processing with dynamic CPU core allocation for each stream, ensuring loss-free data handling at the required 13+ Gbit/s sustained throughput. The solution includes runtime configuration capabilities for metadata handling and camera activation/deactivation, as well as optimized memory management to minimize copy operations and CPU load.

The Result

The GigE Vision driver was successfully deployed in Daimler's vehicle testing infrastructure in cooperation with Bosch. The system achieves loss-free capture at over 13 Gbit/s sustained throughput across all four camera channels. The driver went through multiple release versions (M2, M3, M4, v1.1, v1.2, v1.3) with continuous performance improvements. This collaboration between PLC2 Design GmbH, Daimler AG, and Robert Bosch GmbH resulted in a production-ready solution for automotive ADAS development and validation.

Our Stack