One Stack to Rule them All: To Drive Automated Vehicles, and Reach for the 4th level

TL;DR

The paper addresses the need for a scalable, adaptable autonomous driving stack that can be deployed across different vehicles and sensor configurations. It introduces the TCS-AD architecture with fixed interfaces and interchangeable modules spanning localization, perception, planning, execution, diagnostics, and V2X, enabling rapid testing of new research components. The stack is demonstrated on multiple real vehicles (CoCar and FZI-Shuttles) and in simulation, with extensive real-world demonstrations totaling over 3000 km in urban environments, and validation at multiple test sites. Key contributions include a Lanelet-based HD-map integration, a dual-mode localization strategy, a modular perception pipeline with dynamic/static obstacle handling, PSO-based trajectory planning, safety guards, and V2X capabilities, all designed to support reproducible, cross-vehicle experimentation and collaborative development.

Abstract

Most automated driving functions are designed for a specific task or vehicle. Most often, the underlying architecture is fixed to specific algorithms to increase performance. Therefore, it is not possible to deploy new modules and algorithms easily. In this paper, we present our automated driving stack which combines both scalability and adaptability. Due to the modular design, our stack allows for a fast integration and testing of novel and state-of-the-art research approaches. Furthermore, it is flexible to be used for our different testing vehicles, including modified EasyMile EZ10 shuttles and different passenger cars. These vehicles differ in multiple ways, e.g. sensor setups, control systems, maximum speed, or steering angle limitations. Finally, our stack is deployed in real world environments, including passenger transport in urban areas. Our stack includes all components needed for operating an autonomous vehicle, including localization, perception, planning, controller, and additional safety modules. Our stack is developed, tested, and evaluated in real world traffic in multiple test sites, including the Test Area Autonomous Driving Baden-Württemberg.

Figures (6)

• Figure 1: The TCS-AD stack provides all the capabilities that are needed for driving on public roads. This ranges from localization, perception, and planning to controller. The performance of the software stack has been demonstrated on various types of vehicles and across multiple test environments. \ref{['sec:fzi-shuttles']} which do not have a steering wheel.
• Figure 2: Architecture overview of the TCS-AD function.
• Figure 3: Exemplary visualization for the resolving process for computing the maneuver. In the first step, all intervals are intersected, and the weights are added. The second step includes the search for the equilibrium of the forces. The green interval sets a limit on the maximum extension of the driving area, while the remaining weights vote for an extension, represented in the blue intervals. The consent is found by searching for the change of sign.
• Figure 4: User interface of the diagnostic module. It enables fast debugging of the current state of the AD function. This is especially useful for the safety operator, which do not have deep insight into the driving function.
• Figure 5: All test vehicles of the FZI are presented. They were all equipped at the time of construction with state-of-the-art LiDAR system and D-GPS sensors.