Real-time railway corridor perception

The Australian railway industry contributes an estimated $30 billion to the Australia economy with utilisation expected to increase in the future. Maintenance of both the passenger and freight networks, routinely requires personnel to perform duties from within the rail corridor, a task that holds inherent risk.

This project aims to develop a real-time and light-weight perception unit for use in the rail corridor, enabling remote or semi-autonomous completion of many routine tasks.

Solid-state LiDAR technology will be combined with GPS and neural network solutions to create a device that can detect changes within the rail corridor and track objects of interest. The device itself will be lightweight and require minimal computing resources, allowing it to be incorporated into a variety of existing and emerging platforms.

Participants

• Australasian Centre for Rail Innovation (ACRI)
• University of Newcastle
  

Project background

In the past decades there has been an increasing adoption of GPS and LiDAR technologies for data collection within the rail corridor. Such devices can be mounted onto existing rolling stock and collect data without exposing personnel to the danger zone. In more recent years, advances in neural network (NN) and camera technologies have seen significant R&D into data collection and object detection using such technologies.

Implementation of such systems, however, typically requires significant computing resources locally and post-processing to adequately interpret the large amount of collected data.

This project seeks to combine recent NN technologies with state-of-the-art solid-state LiDAR sensors and classical mapping solutions from robotics to produce a real-time perception solution for use in the rail environment. The approach is to pre-process existing network laser-scan data into grid-maps, enabling the technique of localisation to be utilised. Once localisation has been achieved, the sensing solution will have the capacity to identify changes within the rail corridor and track moving objects such as personnel and animals in real-time.

In contrast to many existing systems, the solution will only require minimal computing resources. The solution itself will be a standalone sensing solution that has the capacity to be applied to a variety of platforms, such as existing rolling stock, measurement trolleys, drones, emerging trackside devices, or level crossing wayside. Some of which may be further demonstrated as part of a final project showcase.

Project objectives

The objective of this project is to develop a self-contained smart sensor that will enable a variety of semi-autonomous tasks within the rail corridor.

The developed sensor should have the following capabilities:

• Construct maps and localise within the rail corridor
• Detect and record significant changes within the rail corridor
• Detect and track objects of interest, such as personnel, within the corridor

Please note …

This page will be a living record of this project. As it matures, hits milestones, etc., we’ll continue to add information, links, images, interviews and more. Watch this space!