Removing Blind Spots

Road traffic has increased tremendously over the past decades, confronting politics and industry with considerable road safety and efficiency challenges. Vehicle-to-vehicle (V2V) communication is generally regarded as one of the main emerging technologies to address these challenges by allowing vehicles to exchange information, about themselves (cooperative awareness) or other detected objects (collective perception) by video, radar or lidar systems. The latter, however, is not only interesting for the V2V link, but can further be used by sensorequipped infrastructure to support vehicular perception over the infrastructure-to-vehicle (I2V) link, opening a whole new range of possibilities. In this paper we first introduce a demonstrator we developed in the scope of the German project 5G NetMobil and prove the high added value of road infrastructure-assisted collective perception for vulnerable road user (VRU) protection. Subsequently, we extend the findings to large-scale scenarios by investigating the performance on highways with different traffic densities in terms of reliability and link type. A key finding is that, in high-density scenarios, the reliability of the I2V link is reduced by the V2V transmissions, so that a more efficient resource allocation scheme maximizing the value of transmitted information is needed to optimize the wireless channel usage. Introduction Mobility is one of the essential drivers of modern times. A world without vehicles is unconceivable for most people nowadays. While the number of vehicles registered worldwide merely exploded in the first one hundred years since Karl Benz’s famous patent in 1886 [1], reaching as much as 440 million units in 1990 [2], that number has more than doubled in the past 30 years [3]. And the global traffic density is expected to increase even further, as (i) the number of vehicles per 1000 inhabitants in less developed countries is still far behind that of their developed peers, and (ii) the world population is expected to continue growing at a rate VDI-Berichte Nr. 2384, 2021 29 https://doi.org/10.51202/9783181023846-29 Generiert durch IP '34.212.30.79', am 04.11.2021, 09:18:35. Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig. of roughly 1% per year in the midterm [4]. This vertiginous development of the traffic density confronts the politics and industry with tremendous challenges when keeping high levels of traffic efficiency and comfort while trying to guarantee road safety. The two probably most promising cutting-edge technologies are traffic automation and vehicle-to-everything (V2X) communication. While they are clearly distinct from each other, the synergies are evident. Intervehicle communication can, e.g., greatly enhance the environmental perception of automated vehicles to strengthen their decision basis, with very positive effects on the mentioned road safety, efficiency, and comfort [5, 6]. Different V2X services have been proposed to extend a vehicle’s perception beyond the perception range of its sensor systems. Cooperative awareness [7] enables road users to share information about their current and past states. It is thus generally very reliable and accurate; however, it lacks sufficient availability until complete V2X market penetration [8]. Collective perception [9] in turn allows road users equipped with object-tracking sensors to share detected objects. While this service does generally not require full V2X market penetration to ensure a complete environmental perception, hence presenting a high availability [10], its reliability and accuracy may be compromised by the environment-dependent quality of their detections [11]. Not only weather conditions [12] but also occlusion by other traffic participants [10] and a limited sensor accuracy [11, 13] may negatively impact the performance of the service. Another service, introduced only recently, is starting to gain attention in the V2X community: collaborative localization [14]. Enabled vehicles can share GNSS and range measurements in order to enhance the localization accuracy and availability of all connected vehicles in the environment as compared to cooperative awareness. However, the availability is still limited to the perception of connected vehicles. Another challenge of deploying V2X services is that their efficient operation depends on the communication performance. Signal attenuation and interference may cause packet losses, compromising the performance of the system [15, 16]. The performance is further sensitively dependent on the absolute positioning capabilities of the connected stations, as shared position data generally needs to be converted to absolute coordinates by the transmitter and back to relative coordinates in the receiver [17]. Infrastructure-assisted connected driving has considerable potential to deal with most of these challenges: (i) road side units (RSUs) can be stationed at strategical positions (sensor hight, angle, distributed arrangement) to guarantee high detection availabilities and accuracies, (ii) these RSU locations may further consider the V2X signal propagation within the area of interest, increasing the communication reliability, and (iii) being static, the RSUs’ absolute positions are well defined, reducing the introduced transformation errors. VDI-Berichte Nr. 2384, 2021 30 https://doi.org/10.51202/9783181023846-29 Generiert durch IP '34.212.30.79', am 04.11.2021, 09:18:35. Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig. In this work we present the following two studies on infrastructure-assisted collective perception: 1. Use case of V2X-based collision avoidance with pedestrians: A roadside infrastructure is equipped with a camera which detect pedestrians crossing the street. These detections are transmitted to oncoming vehicles, which then calculate the collision probability and, if needed, perform an automated braking manoeuvre. This function has been recently demonstrated as a proof-of-concept. 2. Simulation results extend the investigation to large-scale scenarios, proving the benefit of infrastructure-assisted collective perception for elevated numbers of connected stations. The investigation analyses key performance indicators such as reliability and latency in a highway scenario where a roadside unit supports the vehicles’ environmental perception by detecting nearby vehicles and transmitting its data using collective perception. Finally, the next steps to further enhance automated driving using V2X communication, such as a data-quality dependent channel resource allocation are described. Infrastructure-assisted Collective Perception Collective perception allows vehicles and RSUs to inform nearby vehicles of objects (e.g. pedestrians, motorcycles, or other vehicles) detected by their on-board sensors [18]. This enables receiving vehicles to extend their own environmental model beyond their own sensors range by looking through the “eyes” of others. The object data received through V2X communication are then incorporated into the vehicle’s environmental model, eventually combining the object data with that obtained from the on-board sensors by means of data fusion [19, 20]. This approach increases the redundancy of the detected object data [21], leading to a higher reliability. The exchange of object data is done by means of Collective Perception Messages (CPM)s, currently in standardization by the European Telecommunications Standards Institute (ETSI) [9] in order to ensure its interoperability among all equipped vehicles. The term “Infrastructure-assisted Collective Perception (ICP)” refers to the case when the object detection is performed by an RSU, which transmits the object data to nearby vehicles in CPMs [22]. ICP has been the focus of several research projects. The EU-funded TransAID project [23] performed a demonstration of ICP with infrastructure equipped with a video camera using the ETSI CPM. The considered use cases are a blocked road assistant and cooperative VDI-Berichte Nr. 2384, 2021 31 https://doi.org/10.51202/9783181023846-29 Generiert durch IP '34.212.30.79', am 04.11.2021, 09:18:35. Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig. lane merge. The German projects KoMo:Dnext [24] and LUKAS [25] explore the use of ICP to support automated driving at an urban intersection. For this purpose, data for a high-resolution multimodal environment detection are aggregated on the infrastructure side and made available to the surrounding traffic. Equipped road users serve as mobile sensor, transmitter and receiver units and form a multi-sensor network together with the infrastructure. This network generates data to determine the position and dimension of dynamic objects (motorized vehicles, pedestrians, and cyclists) in the observed traffic space in terms of time. These data are transmitted in CPMs and integrated in the vehicles’ Local Dynamic Map (LDM). Fig. 1 shows a sample scenario where ICP allows increasing the traffic safety. Fig. 1: Infrastructure-assisted Collective Perception in an intersection. Vehicle A detects vehicle B with its on-board sensors and transmits the corresponding detected object in a CPM. The RSU receives the CPM, identifies that the message is relevant for vehicle D and that vehicles A and D are not in communication range, since their lineof-sight is blocked by a building. Furthermore, the RSU detects vehicle C with its own sensors and realizes that it is also relevant for vehicle D. Therefore, the RSU retransmits a CPM containing information about vehicles B and C towards vehicle D, which then adapts its maneuver to prevent a collision risk with them. VDI-Berichte Nr. 2384, 2021 32 https://doi.org/10.51202/9783181023846-29 Generiert durch IP '34.212.30.79', am 04.11.2021, 09:18:35. Das Erstellen und Weitergeben von Kopien dieses PDFs ist nicht zulässig. Demonstrator: Vulnerable Road User (VRU) Protection In scope of the Germa