Analysis of Risk in Real-World Driving Scenarios: Implications for Advanced Driving Assistance Systems

Bhowmik, Moumita

Analysis of Risk in Real-World Driving Scenarios: Implications for Advanced Driving Assistance Systems

Files

Primary bhowmik_moumita_msc_2026_thesis.pdf (38.67 MB)

Date

2026-04-02

Authors

Bhowmik, Moumita

Abstract

This thesis presents a dynamic risk assessment algorithm for Advanced Driver Assistance Systems (ADAS) using real urban driving data. The goal is to estimate how risky a current traffic situation is by combining information about the ego vehicle, surrounding objects, and their interactions. To achieve this, a YOLOv8m-DeepSORT pipeline detects and tracks surrounding road users and combines that with CAN bus and IMU signals to obtain ego-vehicle and object states. From these, Time-to-Collision (TTC), Time-to-Materialization (TTM), geometric intersection points, and a set of critical scenarios (crossing traffic, head-on collision, intersection turning, parked/occluded vehicles, not following traffic signs, construction speed-limit violations, and emergency braking) are derived. These inputs feed a Severity–Exposure–Controllability (SEC) model that outputs a continuous risk score R ∈ [0, 10] and LOW, MEDIUM, or HIGH risk classes. The results show realistic scenarios and risk distributions, and a qualitative validation confirms that the proposed SEC-based risk assessment algorithm can provide an interpretable, runtime measure of driving risk that is consistent with established safety concepts and suitable as an input for future ADAS decision-making.

Summary for Lay Audience

This thesis is about making today’s driver-assistance systems better at understanding when a traffic situation is becoming risky. Nowadays, Advanced Driver Assistance Systems (ADAS) can detect cars, pedestrians, and traffic lights, but they mainly react to simple rules (for example, “If something is close, brake!”). They do not always have a clear sense of “How dangerous is this right now?” Safety Standards, on the other hand, are mostly done on paper. Engineers decide in advance which situations are risky and assign safety levels after experimenting with driving simulators. Neither approach actually gives the vehicles a live, moment-by-moment sense of how risky the current scene is. In this work, real driving data on Ottawa roads are used to build a system that looks at the camera view, vehicle sensors, and the movement of surrounding road users to estimate risk frame by frame. It detects typical dangerous situations such as crossing traffic at intersections, head-on approaches, sharp turns, parked cars that hide other road users, and sudden braking. For each situation, it calculates a risk score from 0 (safe) to 10 (very dangerous) and also labels it as low, medium, or high risk. The results show that this score behaves in an intuitive way: it goes up as time to possible collision becomes shorter, as distance gets smaller, and as speed increases. The thesis argues that this kind of continuously updated risk signal could help future driver-assistance systems warn drivers earlier and choose safer actions in complex urban traffic.

Keywords

Advanced Driver Assistance Systems (ADAS), Autonomous Vehicles (AV), Risk Assessment, Risk Scoring, ISO 26262, SEC (Severity–Exposure–Controllability), Safety of the Intended Function (SOTIF, ISO/PAS 21448), Hazard Analysis and Risk Assessment (HARA), Time-to-Collision (TTC), Time-to-Materialization (TTM), Hazard Model, Critical Scenarios, YOLO, Object Detection, DeepSORT, Multi-Object Tracking, Intersection Analysis, Hazard Stages