Interview of Christian Chavanel
UIC Railway System Director

Christian Chavanel, UIC Railway System Director

Christian Chavanel is a railway professional with more than 30 years’ experience in international development, project management, operation, maintenance, safety, standardisation, and regulatory affairs. He is an engineer and holds an Executive MBA. He holds a certicifate from the College of Europe and a certificate from MIT on Artificial Intelligence.He has notably been:

• Interoperability & Standardization Director at SNCF from 2014 to 2019

• Chairman of CEN-CENELEC Sector Forum Rail (ex-JPC-R) from 2016 to 2019

• COO (SNCF Regional Transportation)

• PMO (SNCF Regional Transportation)

• Head of Paris Gare de Lyon station

• Infrastructure District Manager

The UIC Rail System Department supports the work of the Rail System Forum. This forum relies on its members to continuously improve the railway system. The forum is divided into six sectors dedicated to keep railways at the edge of technology and to seamlessly interconnect with other modes of transport. Rail System deals with a wide range of subjects such as Train-Track Interaction, Track and Structures, Rolling Stock, Energy Management, Asset Management and Operations, Telecoms, Signaling and Digital Applications. More than 150 experts are involved in the department’s activities, which covers 78 ongoing projects.

According to you, what means Artificial Intelligence?

The term ‘Artificial Intelligence’ (AI) is a suitcase term and is not easy to define. There is nothing artificial about it. A.I. is made by humans, intended to behave by humans and, ultimately, to impact humans lives and human society. Massive computing power, which only performs requested routine tasks and is controlled at every step by software programmers using a classical analytical approach, is not AI. A classical algorithm, however complicated, which does not deviate from the problem-solving method programmed by a software programmer, is not capable of learning. It is not AI. A large part of artificial intelligence today is based around automatic learning.

Today there are three important subtypes for IA : Machine Learning, Natural Language Processing, and Robotics.

For machine Learning, what are the key points to be focused on?

The typical Machine Learning process requires both data and algorithms. A three-step process maximises the chances of learning success (Towards Data Science, 2018):

• Training: “A subset of real data is provided to the data scientist. The data includes a sufficient number of positive and negative examples to allow any potential algorithm to learn. The data scientist experiments with a number of algorithms before deciding on those which best fit the training data.”

• Validation set: “The data scientist will run the chosen algorithms on the validation set and measure the error. The algorithm that produces the least error is considered the best. It is possible that even the best algorithm can overfit or underfit the data, producing a level of error which is unacceptable.”

• Testing: “To obtain an accurate and reliable measure of error, a third set of data should be used, known as the test set. The algorithm is run on the test set and the error is calculated.” During this learning process, data scientists must choose the best algorithm from a considerable number of them.

The quality of the data chosen is also crucial.

Eight biases must be avoided : propagating the current state, training on the wrong thing, under-representing populations, faulty interpretation, cognitive biases, analytics bias, confirmation bias, and outlier bias (Search Business Analytics, 2020).

What are the business applications for the Railways concerning Natural language processing?

Natural Language Processing (NLP) is at the crossroads of linguistics, computer science and artificial intelligence. It deals with the interactions between computers and human language (text and speech). The result is a computer able to ‘understand’ the content of documents or speeches, including the contextual nuances of the language within them. Existing technology can accurately extract the information and insights contained in documents or in speeches, as well as categorise and organise the documents themselves.

The main application for the Railways is customer service : many companies transcribe and analyse recordings of customer calls. They also deploy chatbots and automated online assistants to provide immediate responses to simple needs and reduce the workload of customer service representatives.

What are the different use cases for robotics?

The different use cases of Robotics are:

• Industrial robots able to streamline picking and packing processes in warehouses.

• Robots facing people: robots as workers in the fast food industry, robots delivering medication in hospitals, robots able to accompany ill people in hospitals in Japan, robots cleaning up airports or railway stations in the presence of passengers, and robots providing passenger information.

• Collaborative robots specifically designed to work alongside human employees are on the rise. They are cheaper, built with human cooperation in mind, and therefore easier to programme.

How AI technologies are currently being deployed within the railway sector?

At this stage, forerunners have carried out innovative solutions. But AI has not yet been widely implemented in Europe.

For example, the current innovative solutions implemented are :

• Image recognition in the fight against terrorism

• Chatbots and virtual assistants for passengers

• Sales prediction through ML

• Robotics in railway stations

• Robotics in trains

• Robotics in warehouses

What are the main perspectives?

The main perspectives for the Railway sector is predictive maintenance for both infrastructure and rolling stock.

What could be the role of UIC regarding AI?

AI technologies solve problems but remain rather opaque (Machine Learning, Deep Learning, Convolutional neural networks, NLP, etc). Consequently, their interpretability is an essential question. However, there is currently no system on the market for interpreting the results provided by Machine Learning. Under these conditions, the role of experts will remain crucial for several years to come. In addition, the authorisation to place AI on the market, particularly for safety cases, will be granted only if the human-machine system as a whole is considered.

UIC will help its Members to create a vision and to publish guidelines for each use case, notably for predictive maintenance for infrastructure and rolling stock.