It is believed that some £40 million per annum is wasted in over-runs on track possessions in both contractor costs and fines paid by Network Rail to the operators for lost service. Given the limited capacity of the existing track infrastructure, we are seeking new ways of maximising capacity to try to improve the efficiency of possessions while improving railway worker safety and reducing delays due to engineering works.
With funding from RRUKA, the MEWOPS (maximising effective working-time on possessions) team led by Professor Muller of University College London’s Mullard Space Science Laboratory, performed a study. This included obtaining more detailed information by interviewing key personnel at Network Rail and its contractors; reviewing relevant technologies, looking at the best possible technologies and methods of data fusion and developing a roadmap for the future.
Railway tracks will, by their very nature, require long-term preventative maintenance and represent by their geography an open system which is near impossible to keep fenced off. Methods for track possession have changed very little since the early 1900’s and rely upon a complex network of procedures, protocols and equipment dating back to the 19th century to ensure safety and security of track workers before, during and after possessions. The key to the current system is the PICOP (person in charge of possessions) who must orchestrate the laying down of detonators (to warn of impending trains), manage the contractors performing the possession and ensure that workers are not placed in harm’s way when trains pass through the possession.
Military systems wed to 3D imaging
The study found that the key to providing a more sustainable solution in the future is to exploit the advances in communications technologies from military systems and wed these to persistent and omnipresent 3D imaging so both the train driver and the PICOP have the best possible information on who is where and what they are doing and what steps can be taken to maximise the safety of everyone involved in the possession.
Communications imply use of either technologies for local point-to-point robust video-rates or an all pervasive internet infrastructure to allow those with the need to know the best possible visual information on the current situation. Experience of members of the team with robotic space exploration, 3D imaging and control and communications technologies greatly assisted the study performance.
The first discovery made in the MEWOPS study is the lack of internet connectivity across the whole track network, either for WiFi or 4G (needed for imaging) or even 3G GSM telecommunications. This poses significant challenges for communicating to the PICOP where potential threats exist (e.g. where are the trains) and how far are they from the possession as well as enabling easy communication on secure links between the PICOP, train driver and signalling and control personnel. This suggests that point-to-point communications will be required so that the train driver is able to communicate over short distances (a few Kms) to the PICOP and vice versa whilst the much slower development of an internet infrastructure happens for the track network.
Providing ‘virtual eyes’ to the driver and PICOP are critical to enhance safety and security of any track possession.
Similar issues in space exploration
Researchers in robotic space exploration have for more than two decades been working on similar issues to ensure the safety of robotic exploration vehicles working in unstructured and hostile environments. A key to this is to provide the driver with visual and auditory inputs on what is happening in front, to the side and behind the train and sharing these views and auditory signals with the PICOP. Everyone is used to having a version of this equipment in their car with rear-viewing cameras and proximity auditory sensors. However, at night and during inclement weather conditions, most motor vehicle vision systems produce noisy and inconclusive results. To address this need, thermal InfraRed imaging needs to fused with 3D range information and ultrasonic systems so that drivers have full situational awareness of the dangers around them whilst communicating this with the PICOP.
Systems have been developed for the coal mining industry that employ RFID detection within a 100m radius so that the PICOP and driver can be notified on an individual basis if anyone tagged is within dangerous proximity of the train. All the command and control aspects of the telecommunications have been thoroughly tested through systems for IFF/AIS (identification of friend or foe/automatic identification system). An existing alternative developed by the German Aerospace Centre, DLR is RCAS, as it seems to provide many of the aspects required by a PICOP advising of oncoming trains.
Robotic systems rely on vision coupled with 3D range information to navigate in complex environments and motor vehicle collision avoidance systems use somewhat similar approaches. Surprisingly, this has not yet happened in the rail industry. The consortium plans to develop such a prototype system and test them in appropriate environments. An example of a range dataset of a typical train environment is shown taken from a terrestrial lidar [Fig.1] while a 3D model of a moving person can be seen in Fig.2 above.
Professor Jan-Peter Muller, UCL Mullard Space Science Laboratory, Blue Sky Imaging Ltd and i3DR Ltd); Nicholas Bantin (IS Instruments Ltd & i3DR Ltd); Tom Jackson (York University & Cybula Ltd)