Recently, the first successful deployment of a fully automated commercial freight train operation was announced. This is the world’s first automated heavy-duty and long-haul rail network. It’s an impressive achievement, but why has it taken so long to achieve this when driverless urban metros have been in operation for more than 50 years? Although urban metros and freight trains are vehicles moved on rails, their operation and environment differ significantly. Metros operate in closed rail networks, while freight trains operate in open rail networks. However, the same taxonomy is often used to classify automation interchangeably in both environments. This paper provides context and an overview of driving automation in freight rail and reviews the existing taxonomies. This paper starts by providing context with an overview of the general process of driving a vehicle by delimiting its different stages. Next, we describe the overall process of driving a freight train to show the distinctive features of its setup and operation. In this analysis, we will point out the essential differences between open and closed rail networks, and the tasks that can potentially be automated. Additionally, we examine the evolution of level-based automation taxonomies and review those that have been proposed exclusively for driving automation in open and closed railway networks. Our objective is to provide a thorough summarization of the most relevant taxonomies to advance the definition of a suitable taxonomy and framework to classify automation capabilities in rail freight transport and identify some complex challenges ahead.

The digitalisation of freight rail is an essential improvement to create modern functions that offer a cost-effective, attractive service and improved operational opportunities to operators. These modern functions need intelligence, detection, actuation and communications. For this, generally, it is possible to process raw data in the Edge and send meaningful data over a communication link. However, the power supply is not granted in a freight wagon and so low power strategies need to be adopted. This paper presents the implementation and testing of a wireless connected heterogeneous multiprocessing architecture. From the power consumption point of view, this system has been stressed by means of a generic FFT function to evaluate the different on-board computing devices that have been decided. From the communication point of view, the LPWAN LoRa technology has been tested and validated on robustness and coverage. Thanks to the heterogeneous nature of this architecture and its configurability, it allows us to propose the most suitable computing ressources, data analysis and communication strategy in terms of efficiency and performance for the functions that this wagon on board unit needs to host and support. With this approach, operation data are reported to the centralised freight driver assistant system.

Air quality degradation in rapidly urbanizing regions involves complex, non-linear dynamics that frequently elude traditional deterministic models. To tackle this limitation, this research introduces a probabilistic forecasting and transfer station methodology. By reliably extracting patterns from official sensor data, our framework predicts the Air Quality Index (AQI) within the Guadalajara Metropolitan Area, Mexico. The approach utilizes NeuralProphet, a hybrid architecture that integrates deep Auto-Regressive Networks with seasonal decomposition. Based on hourly concentrations of six criteria pollutants (${\mathrm{PM}}_{10}$, ${\mathrm{PM}}_{2.5}$, ${\mathrm{O}}_3$, ${\mathrm{NO}}_2$, ${\mathrm{SO}}_2$, $\mathrm{CO}$) collected from four official monitoring stations between 2022 and 2024, a preprocessing pipeline aligned with US EPA standards was implemented. Optimization identified a 12-h lag window to forecast a 1-h future horizon. To evaluate the model’s generalization, the cross-station framework was trained on an aggregated signal from peripheral stations and tested on the industrial core of Miravalle. The framework was benchmarked against standard deep learning architectures (LSTM, GRU, RNN). While recurrent models exhibited high deterministic accuracy, NeuralProphet maintained robust predictive performance ($R^2\approx 0.995$ on the held-out industrial test set) while additionally providing explicit seasonal interpretability and 80% prediction intervals via quantile regression. This quantification of uncertainty helps distinguish stable diurnal trends from sporadic hazardous events. By providing probabilistic forecasting ranges rather than deterministic point estimates, this transparent, data-driven tool offers a practical option for proactive public health interventions alongside black-box alternatives.