Все выпуски

Sensor fusion is one of the important solutions for the perception problem in self-driving cars, where the main aim is to enhance the perception of the system without losing real-time performance. Therefore, it is a trade-off problem and its often observed that most models that have a high environment perception cannot perform in a real-time manner. Our article is concerned with camera and Lidar data fusion for better environment perception in self-driving cars, considering 3 main classes which are cars, cyclists and pedestrians. We fuse output from the 3D detector model that takes its input from Lidar as well as the output from the 2D detector that take its input from the camera, to give better perception output than any of them separately, ensuring that it is able to work in real-time. We addressed our problem using a 3D detector model (Complex-Yolov3) and a 2D detector model (Yolo-v3), wherein we applied the image-based fusion method that could make a fusion between Lidar and camera information with a fast and efficient late fusion technique that is discussed in detail in this article. We used the mean average precision (mAP) metric in order to evaluate our object detection model and to compare the proposed approach with them as well. At the end, we showed the results on the KITTI dataset as well as our real hardware setup, which consists of Lidar velodyne 16 and Leopard USB cameras. We used Python to develop our algorithm and then validated it on the KITTI dataset. We used ros2 along with C++ to verify the algorithm on our dataset obtained from our hardware configurations which proved that our proposed approach could give good results and work efficiently in practical situations in a real-time manner.

Sensor fusion is one of the important solutions for the perception problem in self-driving cars, where the main aim is to enhance the perception of the system without losing real-time performance. Therefore, it is a trade-off problem and its often observed that most models that have a high environment perception cannot perform in a real-time manner. Our article is concerned with camera and Lidar data fusion for better environment perception in self-driving cars, considering 3 main classes which are cars, cyclists and pedestrians. We fuse output from the 3D detector model that takes its input from Lidar as well as the output from the 2D detector that take its input from the camera, to give better perception output than any of them separately, ensuring that it is able to work in real-time. We addressed our problem using a 3D detector model (Complex-Yolov3) and a 2D detector model (Yolo-v3), wherein we applied the image-based fusion method that could make a fusion between Lidar and camera information with a fast and efficient late fusion technique that is discussed in detail in this article. We used the mean average precision (mAP) metric in order to evaluate our object detection model and to compare the proposed approach with them as well. At the end, we showed the results on the KITTI dataset as well as our real hardware setup, which consists of Lidar velodyne 16 and Leopard USB cameras. We used Python to develop our algorithm and then validated it on the KITTI dataset. We used ros2 along with C++ to verify the algorithm on our dataset obtained from our hardware configurations which proved that our proposed approach could give good results and work efficiently in practical situations in a real-time manner.

В работе решается задача установления зависимости потенциала пространственной селекции полезных и мешающих сигналов по критерию отношения «сигнал/помеха» от погрешности позиционирования устройств при диаграммообразовании по местоположению на базовой станции, оборудованной антенной решеткой. Конфигурируемые параметры моделирования включают планарную антенную решетку с различным числом антенных элементов, траекторию движения, а также точность определения местоположения по метрике среднеквадратического отклонения оценки координат устройств. В модели реализованы три алгоритма управления формой диаграммы направленности: 1) управление положением одного максимума и одного нуля; 2) управление формой и шириной главного лепестка; 3) адаптивная схема. Результаты моделирования показали, что первый алгоритм наиболее эффективен при числе элементов антенной решетки не более 5 и погрешности позиционирования не более 7 м, а второй алгоритм целесообразно использовать при числе элементов антенной решетки более 15 и погрешности позиционирования более 5 м. Адаптивное диаграммообразование реализуется по обучающему сигналу и обеспечивает оптимальную пространственную селекцию полезных и мешающих сигналов без использования данных о местоположении, однако отличается высокой сложностью аппаратной реализации. Скрипты разработанных моделей доступны для верификации. Полученные результаты могут использоваться при разработке научно обоснованных рекомендаций по управлению лучом в сверхплотных сетях радиодоступа миллиметрового диапазона пятого и последующих поколений.

The work solves the problem of establishing the dependence of the potential for spatial selection of useful and interfering signals according to the signal-to-interference ratio criterion on the positioning error of user equipment during beamforming by their location at a base station, equipped with an antenna array. Configurable simulation parameters include planar antenna array with a different number of antenna elements, movement trajectory, as well as the accuracy of user equipment location estimation using root mean square error of coordinate estimates. The model implements three algorithms for controlling the shape of the antenna radiation pattern: 1) controlling the beam direction for one maximum and one zero; 2) controlling the shape and width of the main beam; 3) adaptive beamforming. The simulation results showed, that the first algorithm is most effective, when the number of antenna array elements is no more than 5 and the positioning error is no more than 7 m, and the second algorithm is appropriate to employ, when the number of antenna array elements is more than 15 and the positioning error is more than 5 m. Adaptive beamforming is implemented using a training signal and provides optimal spatial selection of useful and interfering signals without device location data, but is characterized by high complexity of hardware implementation. Scripts of the developed models are available for verification. The results obtained can be used in the development of scientifically based recommendations for beam control in ultra-dense millimeter-wave radio access networks of the fifth and subsequent generations.