Accurate 3D trajectory data is crucial for advancing autonomous driving. Yet, traditional datasets are usually captured by fixed sensors mounted on a car and are susceptible to occlusion. Additionally, such an approach can precisely reconstruct the dynamic environment in the close vicinity of the measurement vehicle only, while neglecting objects that are further away. In this paper, we introduce the DeepScenario Open 3D Dataset (DSC3D), a high-quality, occlusion-free dataset of 6 degrees of freedom bounding box trajectories acquired through a novel monocular camera drone tracking pipeline. Our dataset includes more than 175,000 trajectories of 14 types of traffic participants and significantly exceeds existing datasets in terms of diversity and scale, containing many unprecedented scenarios such as complex vehicle-pedestrian interaction on highly populated urban streets and comprehensive parking maneuvers from entry to exit. DSC3D dataset was captured in five various locations in Europe and the United States and include: a parking lot, a crowded inner-city, a steep urban intersection, a federal highway, and a suburban intersection. Our 3D trajectory dataset aims to enhance autonomous driving systems by providing detailed environmental 3D representations, which could lead to improved obstacle interactions and safety. We demonstrate its utility across multiple applications including motion prediction, motion planning, scenario mining, and generative reactive traffic agents. Our interactive online visualization platform and the complete dataset are publicly available at app.deepscenario.com, facilitating research in motion prediction, behavior modeling, and safety validation.

This article considers the problem of diagnosing certain common errors in reward design. Its insights are also applicable to the design of cost functions and performance metrics more generally. To diagnose common errors, we develop 8 simple sanity checks for identifying flaws in reward functions. We survey research that is published in top-tier venues and focuses on reinforcement learning (RL) for autonomous driving (AD). Specifically, we closely examine the reported reward function in each publication and present these reward functions in a complete and standardized format in the appendix. Wherever we have sufficient information, we apply the 8 sanity checks to each surveyed reward function, revealing near-universal flaws in reward design for AD that might also exist pervasively across reward design for other tasks. Lastly, we explore promising directions that may aid the design of reward functions for AD in subsequent research, following a process of inquiry that can be adapted to other domains.

This article considers the problem of diagnosing certain common errors in reward design. Its insights are also applicable to the design of cost functions and performance metrics more generally. To diagnose common errors, we develop 8 simple sanity checks for identifying flaws in reward functions. We survey research that is published in top-tier venues and focuses on reinforcement learning (RL) for autonomous driving (AD). Specifically, we closely examine the reported reward function in each publication and present these reward functions in a complete and standardized format in the appendix. Wherever we have sufficient information, we apply the 8 sanity checks to each surveyed reward function, revealing near-universal flaws in reward design for AD that might also exist pervasively across reward design for other tasks. Lastly, we explore promising directions that may aid the design of reward functions for AD in subsequent research, following a process of inquiry that can be adapted to other domains.

Motion planning for car-like robots is one of the major challenges in automated driving. It requires to solve a two-point boundary value problem (BVP) in real time while taking into account the nonholonomic constraints of the vehicle and the obstacles in the non-convex environment. This paper introduces Hybrid Curvature Rate (HCR) and Continuous Curvature Rate (CCR) Steer: Two novel steering functions for car-like robots that compute a curvature rate continuous solution of the two-point BVP. Hard constraints on the maximum curvature, maximum curvature rate, and maximum curvature acceleration are satisfied resulting in directly driveable G3continuous paths. The presented steering functions are benchmarked in terms of computation time and path length against its G1and G2continuous counterparts, namely Dubins, Reeds-Shepp, Hybrid Curvature, and Continuous Curvature Steer. It is shown that curvature rate continuity can be achieved with only small computational overhead. The generic motion planner Bidirectional RRT* is finally used to present the effectiveness of HCR and CCR Steer in three challenging automated driving scenarios.

In the near future, humans will be relieved from parking. Major improvements in autonomous driving allow the realization of automated valet parking (AVP). It enables the vehicle to drive to a parking spot and park itself. This paper presents a review of the intelligent vehicles literature on AVP. An overview and analysis of the core components of AVP such as the platforms, sensor setups, maps, localization, perception, environment model, and motion planning is provided. Leveraging the potential of AVP, high density parking (HDP) is reviewed as a future research direction with the capability to either reduce the necessary space for parking by up to 50 % or increase the capacity of future parking facilities. Finally, a synthesized view discussing the remaining challenges in automated valet parking and the technological requirements for high density parking is given.

Advances in autonomous driving and the introduction of automated valet parking allow the optimization of parking space. A future concept is high density valet parking with the potential to either reduce the extensive land use for parking or increase the capacity of existing parking facilities. This paper presents a novel approach that integrates high density parking into an existing parking lot, by explicitly making use of parking in the driving lane and reducing the shunting operations per vehicle. The proposed parking scheme allows vehicles to park either perpendicular to the driveway or in double-ended queues with k parking spots (k-deque) on the side of the driving lane. Leveraging the potential of such a layout increases the capacity of a parking lot by up to 25 %, while keeping the maximum number of shunts per car below [k/2] +1 between entry and exit. A dynamic simulation verifies the theoretical analysis and compares the performance of different deque lengths with respect to the distances traveled and the number of shunts per vehicle.