Braking Curves in Railway Shunting and Implications for the Development of Sensor Systems for Autonomous Shunting

Shunting operations differ largely from train operation on the mainline in both train protection and braking. While for mainline operations, braking of all axles in a train set is mandatory under most circumstances and unbraked wagons and axles are an exception, common shunting regulations allow for up to 40 unbraked axles depending on the locomotive mass and track gradient. On the other hand, train protection in shunting operation is mostly achieved on an on sight basis, which means that no technical devices ensure the freedom of movement. It is rather common to discuss automatic train operation (ATO) these days, however most discussions focus on the mainline portion of freight rail. In mainline operation however, the movement authority is ensured by signalling, so no long range scanning equipment, such as cameras, radar and LiDAR is strictly required to ensure safety of operation. Also, based on the braking behaviour of the train in question, a rather precise velocity recommendation is provided, either by static signals (brake tables, signage) or by continuous communication (e.g. LZB, ETCS). The opposite holds true for shunting: in most cases, a movement authority is not given or ensured by technical means, rather the observation of the shunting area by the driver and potential assistants check whether the intended track is free and safe. At the same time, the braking behaviour of the shunting groups is not precisely known and predicted by the driver based on instinct and experience. For this reason, to assist and eventually replace the aging workforce of shunting drivers and shunting assistants, a perception system is considered more demanding since it cannot rely on ATP infrastructure as in the mainline case. In this paper, the special brake setup for shunting mode is analysed and braking curves for numerous cases, ranging from an individual locomotive using direct brake only to a train consist with the maximum number of unbraked axles, are simulated. The simulation software takes into account the train setup, braked weight and brake mode as well as variations in the loading state and the friction and adhesion parameters. The corresponding braking distances are inspected and put into relation to common track geometries and use cases in shunting areas. Further, the visibility requirements from the respective European regulations are reviewed based on these use cases. The requirement set made up of track geometries, visibility and the respective braking curves provides the input for the generation of a set of requirements for the development of a perception system for shunting operation. Example data from current tests of a candidate sensor system will be shown.