Research Projects

Motorcycle crashes account for a disproportionately high number of road traffic fatalities in India due to their inherent instability and minimal protective features. While Advanced Driver Assistance Systems (ADAS) have enhanced vehicle occupant safety worldwide, equivalent solutions for motorcycles remain underdeveloped, especially in the Indian context. This project proposes the development of an Advanced Rider Assistance System (ARAS), tailored for Indian traffic environments, leveraging behavioral analytics and trajectory-based prediction techniques. The project is structured around four interlinked objectives. In the first phase, high-resolution video data will be collected from varied road settings, including intersections, mid-blocks, and roundabouts across urban and semi-urban areas. Using advanced computer vision algorithms, trajectory data of motorcycles, pedestrians, and vehicles will be extracted. From this data, critical motion variables, such as speed, acceleration, lateral spacing, and surrogate safety metrics like Time-To-Collision (TTC), will be computed. Trained observers will classify safety-critical interactions and identify causal factors, distinguishing rider-induced behaviors (e.g., weaving, abrupt lane changes), external influences (e.g., pedestrian unpredictability), and environmental hazards (e.g., poor visibility, damaged roads). Additionally, behavioral pattern mining will uncover latent high-risk riding clusters that manual review may overlook. In the second phase, the annotated trajectory data will inform the development of predictive models. These models will estimate the probability and severity of upcoming conflicts using machine learning techniques that integrate rider motion features and contextual traffic parameters. The aim is to anticipate hazardous maneuvers or collisions several seconds in advance, enabling proactive safety responses. The third phase will focus on designing a modular, cost-effective ARAS prototype. It will integrate a forward-facing camera, IMU, and GPS module, with edge-computing capabilities for real-time processing. Alerts will be delivered via multimodal feedback, haptic (handlebar vibrations), visual (mirror-mounted LEDs), and auditory cues. An associated mobile app will provide post-ride safety analytics, critical event logs, and violation alerts. Finally, the ARAS system will undergo field trials involving volunteer riders. Trials will start in silent mode (logging predictions without alerts), followed by an active mode (delivering real-time warnings). Rider responses, behavioral changes, and usability feedback will guide refinement. An anonymized dataset of interactions and performance will be shared publicly to support future research and policymaking. Outcomes include peer-reviewed publications, patent filings, and recommendations for large-scale ARAS deployment to improve motorcycle safety in India.