Before Christmas we called for machine learning suppliers with an interest in the rail industry to provide their perspective on the current climate of ML in rail and what the future will hold for the industry. We received some really interesting responses from companies who had engaged all across the rail supply chain. Here’s what we found out.
Most companies which responded to us were relatively young (2-5 years old). They were all already engaged in at least one other industry, most commonly manufacturing, but automotive and aerospace were also popular responses.
When asked what they could offer the rail industry, suppliers cited many of the applications of this technology we have already described in this series, such as asset management, improving safety, and improving the customer experience. One company reported they were developing scalable mobile data solutions to match rural and city performance which would improve coverage across the rail network. Other interesting applications in development included intrusion detection devices and behavioural analytics. These could have many valuable applications for the industry: trespasser identification, crowd flow analysis, identifying customers who are drunk, have injured themselves, or require other assistance could improve safety and enhance the customer experience. We also asked suppliers what products they planned to develop in the future. Many intended to move into new niches within fast developing technological fields, such as 5G, augmented reality, and new battery technologies. This is testament to the key strengths of machine learning: it’s flexibility and adaptability to novel problems, which enables its suppliers to move quickly into new and promising markets.
With such a range of ML products available, suppliers see its impacts across the rail sector. When asked, “where could ML create the most radical change in rail?”, responses ranged from improvements to train control and signalling systems to improved cyber security, passenger flow, and predictive maintenance. The consensus appears to be that wherever there is data, there can be significant change. And rail has a lot of data.
So what do suppliers need from us, that we can jointly realise this optimised, automated future together? The survey painted a mixed picture of rail as a customer. One supplier reported the industry shows appetite to use ML to optimise business operations. Others described rail as generally interested, but reserved about where to safely deploy ML and slow to allocate budgets. Lack of engagement by industry with computer scientists was mentioned several times. Interestingly, no suppliers suggested rail lacked the resources to make use of machine learning. There may be a misconception that to use machine learning, a fleet of staff data scientists are required. However, the majority of these suppliers produce products which require a low level of training and support (we asked them to rate their products training/support requirements on a scale of 1-5, and over 63% gave an answer of 1 or 2).
Suppliers also advise rail to improve its data practises. As we have seen throughout this series, having the right data is the key to harnessing the power of machine learning. In rail we generate vast volumes of data, but according to survey respondents, we will get more use from it if we work to improve its quality, interoperability and availability. This opportunity features in the
Rail Sector Deal and the benefits of making rapid progress on the data front are significant for the customer experience and beyond.
This survey shows that suppliers share our belief that ML has strong potential to improve many aspects of operation within the rail industry. We ourselves, as an industry, have some way to go to create an environment where machine learning can flourish. We need to act now by becoming more receptive to learning about ML (hopefully this blog is a step towards that goal). We need to take steps to make our data more available and more useful. We can also see from these responses that the supplier sentiment echoes our own conclusion from this blog series’ last article: we need to upskill or attract talent to harness machine learning. We need to start doing this now to create an understanding and appreciation of machine learning within our organisations. The building blocks for an intelligent and adaptive, machine learning powered railway are already out there, operating in other industries. The drive to adapt them for the rail market exists in the supplier space. What is needed is our expertise in the workings of the rail industry to shape those adaptations and ensure they are directed towards creating real value for rail.