Recently, autonomous driving systems have attracted a significant amount of interest and in the last decade, noteworthy progress has been made in autonomous driving. Development is pushed by companies like Alphabet (Google), Tesla, and Uber while being chased by conventional automotive companies. The ever-increasing computing power and the progress that has been made in image recognition and understanding enables even further growth in the area.
An autonomous driving system enables a vehicle to drive by itself without human input. It perceives and analyzes the environment and its own position, follows rules like traffic regulations, and is able to reach any specified location safely. Making human drivers obsolete will eliminate mistakes made by human drivers, yet might introduce mistakes made by autonomous driving systems. Expectations are that driving will become safer as a result.
Such a system requires perception, localization, planning, and control, and must include a vast variety of sensors, actuators, and computing units. However, making the "correct" decision based on sensor input at all times is still considered very difficult. Fully autonomous vehicles are therefore still in an experimental phase and the topic of current research.
With image processing, understanding and sensor interpretation being an important part of an autonomous driving system, companies like Alphabet that already deploy algorithms for data analysis at scale can use their advantage and experience in areas such as advanced machine learning and computer vision to develop autonomous driving systems.
In the project group "Intelligente Autonome Taxis" at TU Dortmund we aim to develop such an autonomous driving system to work not only in a simulated environment but also with a physical model car. This task connects multiple different fields such as system architecture, machine learning, computer vision, control theory and validation and also requires us to apply project management techniques. The connection of different fields and to work on something that is part of current research motivates our group the most. But we are also excited to be able to lay the groundwork and make architectural decisions that will be used by future project groups working with the physical cars.
This report and the video describe the development and the architecture of the autonomous driving system and summarizes the results and experience of the project group’s work in the course of a year.