Search

We demonstrate a working functional prototype of a cooperative perception system that maintains a real-time digital twin of the traffic environment, providing a more accurate and more reliable model than any of the participant subsystems—in this case, smart vehicles and infrastructure stations—would manage individually. The importance of such technology is that it can facilitate a spectrum of new derivative services, including cloud-assisted and cloud-controlled ADAS functions, dynamic map generation with analytics for traffic control and road infrastructure monitoring, a digital framework for operating vehicle testing grounds, logistics facilities, etc. In this paper, we constrain our discussion on the viability of the core concept and implement a system that provides a single service: the live visualization of our digital twin in a 3D simulation, which instantly and reliably matches the state of the real-world environment and showcases the advantages of real-time fusion of sensory data from various traffic participants. We envision this prototype system as part of a larger network of local information processing and integration nodes, i.e., the logically centralized digital twin is maintained in a physically distributed edge cloud.

A spectacular measurement campaign was carried out on a real-world motorway stretch of Hungary with the participation of international industrial and academic partners. The measurement resulted in vehicle based and infrastructure based sensor data that will be extremely useful for future automotive R&D activities due to the available ground truth for static and dynamic content. The aim of the measurement campaign was twofold. On the one hand, road geometry was mapped with high precision in order to build Ultra High Definition (UHD) map of the test road. On the other hand, the vehicles—equipped with differential Global Navigation Satellite Systems (GNSS) for ground truth localization—carried out special test scenarios while collecting detailed data using different sensors. All of the test runs were recorded by both vehicles and infrastructure. The paper also showcases application examples to demonstrate the viability of the collected data having access to the ground truth labeling. This data set may support a large variety of solutions, for the test and validation of different kinds of approaches and techniques. As a complementary task, the available 5G network was monitored and tested under different radio conditions to investigate the latency results for different measurement scenarios. A part of the measured data has been shared openly, such that interested automotive and academic parties may use it for their own purposes.

Currently the state-of-the-art monocular 3D object detectors use machine learning to estimate the 6DOF pose and shape of vehicles. This requires large amounts of precisely annotated 3D data for the training process and significant computing power for inference. Alternatively, there exist methods, which attempt to reconstruct target vehicle shapes and scales using projective geometry and classically detected feature points such as SURF and ORB. These methods use specific camera motion or geometrical constraints which cannot always be assumed. The resulting model is an unstructured point cloud which contains no semantic information, making its utility inconvenient in a distributed perception system. In this study, the applicability of semantic keypoints for vehicle shape and trajectory estimation is explored. A novel method is presented, which is capable reconstructing the semantic shape and trajectory of the target vehicle from a sequence of images with state-of-the art accuracy. The resulting semantic vertex model is then used for monocular, single frame 6DOF pose estimation with high accuracy. Building on this, a cooperative perception framework is also introduced. The algorithm is tested in both in-vehicle and infrastructure mounted mono-camera sensor setups. In addition to achieving state of the art depth accuracy in vehicle trajectory reconstruction on the Argoverse dataset, our method outperforms the state of the art shape-aware deep learning method in pose estimation in a cooperative perception scenario both in simulation and in real-world experiments.

In autonomous cars accurate and reliable detection of objects in the proximity of the vehicle is necessary in order to perform further safety critical actions which depend upon it. Many detectors have been developed in the last few years, but there is still demand for more reliable and more robust detectors. Some detectors rely on a single sensor, while some others are based upon fusion of data from multiple sources. The main aim of this paper is to show how image features can contribute to performance improvement of detectors which rely on pointcloud data only. In addition it will be shown, how lidar reflectance data can be substituted by low level image features without degrading the performance of detectors. Three different approaches are proposed to fuse image features with point cloud data. The extended networks are compared with the original network and tested on a well-known dataset and on our own data, as well. This might be important when the same pretrained model is to be used on data generated by a lidar using different reflectance encoding schemes and when due to the lack of training data retraining is not possible. Different augmentation techniques have been proposed and tested on the KITTI dataset as well as on data acquired by a different lidar sensor. The networks augmented with image features achieved a recall increase of a few percent for occluded objects.

The representation of objects in LiDAR point clouds is changed as the height of the mounting position of sensor devices gets increased. Most of the available open datasets for training machine learning based object detectors are generated with vehicle top mounted sensors, thus the detectors trained on such datasets perform weaker when the sensor is observing the scene from a significantly higher viewpoint (e.g. infrastructure sensor). In this paper a novel Automatic Label Injection method is proposed to label the objects in the point cloud of the high-mounted infrastructure LiDAR sensor based on the output of a well performing “trainer” detector deployed at optimal height while considering the uncertainties caused by various factors described in detail throughout the paper. The proposed automatic labeling approach has been validated on a small scale sensor setup in a real-world traffic scenario where accurate differential GNSS reference data where also available for each test vehicle. Furthermore, the concept of a distributed multi-sensor system covering a larger area aimed for automatic dataset generation is also presented. It is shown that a machine learning based detector trained on differential GNSS-based training dataset performs very similarly to the detector retrained on a dataset generated by the proposed Automatic Label Injection technique. According to our results a significant increase in the maximum detection range can be achieved by retraining the detector on viewpoint specific data generated fully automatically by the proposed label injection technique compared to a detector trained on vehicle top mounted sensor data.