Multimodal two-wheeler driving dataset for autonomous driving applications

Abstract Autonomous vehicles are shaping the future of human mobility, with most advancements focused on self-driving cars and other four-wheeled vehicles. However, two-wheelers, such as motorcycles, present unique challenges due to their lack of inherent stability and complex driving dynamics. High-quality data is crucial for the research and development of intelligent systems for autonomous driving. Such data also facilitates the design and simulation of realistic driving scenarios—both from the rider’s perspective and for understanding and predicting the behavior of other road users, including pedestrians and vehicles. These interactions tend to be less structured and more dynamic in developing countries like India. In this paper, we present the collection and processing of IndiGo, a large-scale, multi-modal dataset captured from a two-wheeler in India. The dataset includes high-resolution imaging, panoramic 3D point clouds, GNSS data, velocity measurements, driver’s point-of-view imaging, and IMU data from multiple locations on the vehicle, including the rider’s helmet. This dataset supports a wide range of applications, including 3D object tracking, physically accurate simulations, pedestrian and vehicle behavior prediction, and the development of Autonomous Driving Systems (ADS) and Advanced Driver Assistance Systems (ADAS) for two-wheelers and other vehicles. Through this work, we introduce the most extensive multi-modal dataset available for research in autonomous driving and traffic behavior analysis under unstructured traffic conditions in India. Additionally, we discuss sensor selection, data collection methodologies, post-processing techniques, and data-handling tools.