Autonomous vehicles need to be designed to abide by the same rules that humans follow. This is challenging, because traffic rules are fuzzy and not well defined, making them incomprehensible to machines. Satisfaction cannot be incorporated in a planning component without proper formalization, nor can it be monitored and verified during simulation or testing. However, no research work has provided a consistent set of machine-interpretable traffic rules for a given operational driving domain. In this paper, we propose a methodology for the legal study and formalization of traffic rules in a formal language. We use Linear Temporal Logic as a formal specification language to describe temporal behaviors, capable of capturing a wide range of traffic rules. We contribute a formalized set of traffic rules for dual carriageways and evaluate the effectiveness of our formalized rules on a public dataset.

Predicting the trajectories of other road users relies to a large extent on the assumption that they adhere to the legally binding traffic rules. Hence, when this assumption does not hold anymore, the prediction becomes invalid, putting autonomous vehicles relying on such predictions in a critical situation. We propose a solution to this problem by predicting traffic rule violations. All traffic rules are modeled by temporal logic, and we provide real-valued generalizations of required logical predicates to obtain features for prediction with neural networks. The usefulness of our approach is demonstrated by predicting rule violations on a dataset recorded from a highway. Our results show that directly learning traffic rule violations using the features from temporal logic formulas often performs better compared to separately predicting and monitoring trajectories.

Real-world datasets facilitate the development of autonomous vehicles, especially when they are accessible, diverse, and provide a measure of accuracy. While existing datasets have been accessible and diverse, they cannot provide any measure of accuracy. To estimate the accuracy of the detection of traffic participants in our setup, we repetitively drove through our observation area with a measurement vehicle with highly accurate localization and LiDAR sensors. Our experiments showed an average overall position accuracy of 0.51 m. The combined data of the autonomous vehicle and the elevated camera setup yield a unique dataset. The elevated view acts as a super sensor of the autonomous vehicle with extended range and reduced occlusions. We employ an auto-labeling system on the stationary camera data to extract trajectories with bounding boxes for each traffic participant. These extracted trajectories are smoothed for kinematic feasibility, and corresponding maps for each location are provided. The Munich Motion Dataset of Natural Driving (MONA) shall empower new research in prediction and planning. Making raw data and code available to the public without license restrictions allows the dataset to be further improved using more advanced algorithms.

Autonomous vehicles need to abide by the same rules that humans follow. Some of these traffic rules may depend on multiple agents or time. Especially in situations with traffic participants that interact densely, the interactions with other agents need to be accounted for during planning. To study how multi-agent and time-dependent traffic rules shall be modeled, a framework is needed that restricts the behavior to ruleconformant actions during planning, and that can eventually evaluate the satisfaction of these rules. This work presents a method to model the conformance to traffic rules for interactive behavior planning and to test the ramifications of the traffic rule formulations on metrics such as collision, progress, or rule violations. The interactive behavior planning problem is formulated as a dynamic game and solved using Monte Carlo Tree Search, for which we contribute a new method to integrate history-dependent traffic rules into a decision tree. To study the effect of the rules, we treat it as a multi-objective problem and apply a relaxed lexicographical ordering to the vectorized rewards. We demonstrate our approach in a merging scenario. We evaluate the effect of modeling and combining traffic rules to the eventual compliance in simulation. We show that with our approach, interactive behavior planning while satisfying even complex traffic rules can be achieved. Moving forward, this gives us a generic framework to formalize traffic rules for autonomous vehicles.