Conceptual design and evaluation of sensor technology for automatic uncoupling of railroad wagons

Across Europe the obsolete screw-couplers, which are currently used to connect cargo train wagons to each other, are to be replaced by a new type, the so-called Digital Automatic Coupler (DAC). This new coupling type automates the coupling between two wagons, but, depending on the final implementation, not the decoupling, which still has to be done manually. As such, to make use of the full potential of this new technology and to make a step towards a fully autonomous rail system, the decoupling is to be automated as well within marshalling yards. The purpose of this paper is to develop and evaluate a sensor system for a track-side device that can autonomously perform this task.
Concepts for such decoupling devices have been around for decades, both for centre-buffer-couplings and the obsolete screw-coupler, yet only few of these were ever tested and none found any commercial adoption. This seems to be due to the difficulty of developing such a technical solution and the cost of implementing it.
Therefore, a new sensor system is to be developed, which can perform reliably even in the harsh environments of marshalling yards. The system is to identify and localise couplings which are to be separated. Three different concepts to achieve this are designed and evaluated. The best concepts regarding the results of this evaluation are a distance measurement concept, using laser and radar distance sensors, and a wagon localisation concept, using axle counters. The concept utilising the distance measurement is then built and tested under realistic conditions in a small marshalling yard on a train equipped with DACs.
Using the recorded data from these tests, methods for signal processing are tested and compared to localise the couplings in the datasets. The methods best suited for this regarding the results of this evaluation are a histogram analysis and a cross correlation. An algorithm combining these two methods is developed, which allows for reliable localisation of the couplings on a passing train.
Due to issues with the used sensors the accuracy of this localisation is severely lacking. While the concept of the distance measurement and the algorithm both work well, the used hardware in its used configuration proved to have severe issues. The laser sensors are unreliable in the harsh conditions of the marshalling yard, while the radar sensors are not dynamic enough. However, these issues seem to be resolvable with more effort or using different hardware. Regardless, the results of this paper pose an excellent basis for the future development of this system, which involves improving the sensor system with the experiences gathered up to now and developing the actuator system for interacting with the couplings.