Safety assessment of autonomous vehicles using simulations

• Main Author: Ketki, Chaudhary
• Other Authors: Justin Dauwels
• Format: Theses and Dissertations
• Language: English
• Published: 2018
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: http://hdl.handle.net/10356/76332
• Institution: Nanyang Technological University

Since the past few decades, researchers from both automotive and technology industries have focussed their attention on developing vehicles that could drive themselves, so that travel would be less challenging, free from human driver errors and the roads safer for everyone. Self-driving cars are envisioned as robots capable of precise driving, never getting distracted, abiding to all the traffic rules and being able to compute the probability of an imminent collision well in advance. Regardless of all these safety precautions, there have been instances of crashes involving big autonomous vehicles developers like Tesla, Uber etc. Undoubtedly, this raises a question of how safe autonomous vehicles are and what are the available methods to assess their safety. The work in this thesis focuses on developing a simulation methodology that can potentially evaluate the safety of highly autonomous vehicles (Level 5) in complex traffic environments. The scope of this thesis is limited to safety analysis of a scenario that involves autonomous obstacle avoidance during a lane changing and highway exit manoeuvre, in the presence of high-paced highway traffic. The purpose of this work is to adopt the minimum spacing requirements for both longitudinal and lateral dynamics of the autonomous vehicle. These requirements are derived using an automotive standard-ISO26262 as a reference and each test scenario is analysed for its compliance to the safe limits. Safety is analysed by recording the vehicle behaviour data, using simulations, and calculating some crucial parameters. An initial trajectory (mission planning) is first planned and whenever an obstacle is detected by the AV, the new set of waypoints (path planning) are computed iteratively. For different configurations of the afore-mentioned, the simulation data is analysed using MATLAB and checked whether or not it meets safety constraints. The thesis is concluded with discussions on the results and possible future extension of the work carried out in this thesis.