In this paper, we present techniques for automatically classifying players and tracking ball movements in basketball game video clips under poor conditions where camera angle dynamically shifts and changes. In the core of our system lies Yolo, a realtime object detection system. Given the ground truth boxes collected by our data specialists, Yolo is trained to detect the presence of objects in every video frame. In addition, Yolo uses Darknet that implements Convolution Neural Networks to classify a detected object to a player and to recognize its jersey numbers of specific movements. By identifying players and ball possessions, we can automatically compute ball distributions that are reflected on complex networks. With original Yolo system, player movement can be interrupted when the players move out of the frame due to camera shift and when players overlap each other on a two-dimensional frame.We have adapted Yolo to keep track of players even under such poor condition by taking into account contextual information available from the framework preceding and/or succeeding problematic video frames. In addition to the novel movement inference method, we provide a framework for analyzing the pass networks in various perspectives to help the managing staff to reveal critical determinants of team performance and to design better game strategies. We assess the performance of our system in terms of accuracy by making a comparison with the analytical reports generated by human experts.

Automated Vehicles (AVs) are attracting attention as a safer mobility option thanks to the recent advancement of various sensing technologies that realize a much quicker Perception–Reaction Time than Human-Driven Vehicles (HVs). However, AVs are not entirely free from the risk of accidents, and we currently lack a systematic and reliable method to improve AV safety functions. The manual composition of accident scenarios does not scale. Simulation-based methods do not fully cover the peculiar AV accident patterns that can occur in the real world. Artificial Intelligence (AI) techniques are employed to identify the moments of accidents from ego-vehicle videos. However, most AI-based approaches fall short in accounting for the probable causes of the accidents. Neither of these AI-driven methods offer details for authoring accident scenarios used for AV safety testing. In this paper, we present a customized Vision Transformer (named ViT-TA) that accurately classifies the critical situations around traffic accidents and automatically points out the objects as probable causes based on an Attention map. Using 24,740 frames from Dashcam Accident Dataset (DAD) as training data, ViT-TA detected critical moments at Time-To-Collision (TTC) ≤ 1 s with 34.92 higher accuracy than the state-of-the-art approach. ViT-TA’s Attention map highlighting the critical objects helped us understand how the situations unfold to put the hypothetical ego vehicles with AV functions at risk. Based on the ViT-TA-assisted interpretation, we systematized the composition of Functional scenarios conceptualized by the PEGASUS project for describing a high-level plan to improve AVs’ capability of evading critical situations. We propose a novel framework for automatically deriving Logical and Concrete scenarios specified with 6-Layer situational variables defined by the PEGASUS project. We believe our work is vital towards systematically generating highly reliable and trustworthy safety improvement plans for AVs in a scalable manner.

Automated Vehicles (AVs) based on a collection of advanced technologies such as big data and artificial intelligence have opened an opportunity to reduce traffic accidents caused by human drivers. Nevertheless, traffic accidents of AVs continue to occur, which raises safety and reliability concerns about AVs. AVs are particularly vulnerable to accidents on urban roads than on highways due to various dynamic objects and more complex infrastructure. Several studies proposed a scenario-based approach of experimenting with the response of AVs to specific situations as a way to test their safety. Reliable and concrete scenarios are necessary to test AV safety under critical conditions accurately. This study aims to derive a typical accident scenario for evaluating the safety of AVs, specifically in urban areas, by analysing collisions reported by the DMV of California, USA. We applied a hierarchical clustering method to find groups of similar reports and then executed association rule mining on each cluster to correlate between accident factors and collision types. We combined statistically significant association rules to constitute a total of 14 scenarios that are described according to an adapted PEGASUS framework. The newly obtained scenarios exhibit significantly different accident patterns than the typical Human-driven Vehicles (HVs) in urban areas reported by National Highway Traffic Safety Administration. Our discovery urges AV safety to be tested reliably under scenarios more relevant than the existing HV accident scenarios.

Traffic speed prediction is an important topic in Intelligent Transport Systems (ITS). Although traffic speed prediction for real-life applications is burgeoning, the study of explaining and interpreting AI-based speed prediction is still in its initial stage. In this paper, we applied multiple advanced regression techniques, such as XGBoost and CatBoost optimized gradient boosting, Random Forest, and LASSO to predict traffic speed more accurately in the subsequent time windows. The experiment with prediction methods was conducted using the traffic speed data of the Seoul metropolitan road network. Each road segment represented as a node in the network is associated with neighboring roads within a configurable range. We picked heavily congested nodes as prediction targets. Then, we evaluated nearby road influences to determine critical contributions to the situation of the target nodes. We interpreted the model’s output and extracted the topmost influential neighboring nodes by using an ensemble of explainable artificial intelligence (XAI) techniques such as feature importance assessment using the GINI entropy function, Recursive Feature Elimination, Shapely Additive Explanation, and a method of measuring the impact of masked nodes. We validated the XAI interpretations through traffic flow simulation by tuning the topmost influential nearby roads’ speed and observing the effect on the roads’ traffic congestion relief correspondingly. We also proved our solution through local explanation techniques such as Local Interpretable Model-Agnostic Explanations. Our methods are applicable to any transport network and open the door to new strategies for controlling the specific nearby roads for effective congestion relief.