Refine
Document Type
Language
- English (5)
Publication reviewed
- begutachtet (5)
Keywords
- Frequency Shift Keying (1)
- Internet of Things (1)
- LiDAR sensor (1)
- Low-Complexity Design (1)
- Low-Power Wide Area Networks (1)
- Optimization (1)
- Power Line Communication (1)
- Receiver Synchronization (1)
- Smart Meter (1)
- Wakeup Receiver (1)
- Wireless Network (1)
- advanced driver assistance systems (1)
- deep learning (1)
- highly automated driving (1)
- motion distortion effect (1)
- point cloud (1)
- velocity estimation (1)
Many modern automated vehicle sensor systems use light detection and ranging (LiDAR) sensors. The prevailing technology is scanning LiDAR, where a collimated laser beam illuminates objects sequentially point-by-point to capture 3D range data. In current systems, the point clouds from the LiDAR sensors are mainly used for object detection. To estimate the velocity of an object of interest (OoI) in the point cloud, the tracking of the object or sensor data fusion is needed. Scanning LiDAR sensors show the motion distortion effect, which occurs when objects have a relative velocity to the sensor. Often, this effect is filtered, by using sensor data fusion, to use an undistorted point cloud for object detection. In this study, we developed a method using an artificial neural network to estimate an object’s velocity and direction of motion in the sensor’s field of view (FoV) based on the motion distortion effect without any sensor data fusion. This network was trained and evaluated with a synthetic dataset featuring the motion distortion effect. With the method presented in this paper, one can estimate the velocity and direction of an OoI that moves independently from the sensor from a single point cloud using only one single sensor. The method achieves a root mean squared error (RMSE) of 0.1187 m s−1 and a two-sigma confidence interval of [−0.0008 m s−1, 0.0017 m s−1] for the axis-wise estimation of an object’s relative velocity, and an RMSE of 0.0815 m s−1 and a two-sigma confidence interval of [0.0138 m s−1, 0.0170 m s−1] for the estimation of the resultant velocity. The extracted velocity information (4D-LiDAR) is available for motion prediction and object tracking and can lead to more reliable velocity data due to more redundancy for sensor data fusion.
In this article, the optimization of the control circuit and path planning of a delta kinematic with the help of machine learning is presented. The described delta kinematic is primarily used for pick-and-place applications in the field of packaging machines. The optimization of the path planning procedure aims to make the workflow more efficent and flexible for commissioning the delta kinematic. By optimizing the control circuit using machine learning, mechanical oscillations and the deviation of the specified path are to be minimized. The possible use of a simulator for training, the prediction quality and the implementation on the robot controller are discussed. Furthermore, the path planning procedure was optimized. For this purpose, an environment was implemented in which a reinforcement learning agent plans the path of the robot between a starting point and a target point in a time-optimized manner, considering interference contours e.g. from the machine. The obtained results show the optimization of the robot by machine learning with a root mean squared error of the predicted torques of 0.06025 Nm in a prediction time of around 0.125 ms and the possibility of path planning with different criteria.
Synchronization Approaches and Improvements for a Low-Complexity Power Line Communication System
(2019)
This paper presents several improvements of the energy-pattern based sequence detection (EPSD) algorithm for FSK-based single-phase power line communication (PLC) systems, in terms of complexity, reliability and synchronization. A time synchronization is presented which fulfills the well known task of synchronizing transmitter and receiver, but also helps to avoid transmissions in periods of rough noise conditions. The synchronization method is based on a maximum-likelihood approach that makes use of the phase of the mains voltage. Further improvements concern the codes used for the transmitted sequences as well as the combination of the information within both FSK carrier-frequencies in terms of equal gain and maximum ratio combining. Additionally an approach for a low complexity frame synchronization is presented.
In this article, a system for speed estimation of vehicles in road traffic is presented. Using a state-of-the-art Convolutional Neural Network (CNN) for object detection, vehicles are first recognized as objects in the image material captured by a monocular camera, e.g. a mobile phone. In order to prevent the fluctuations of the bounding boxes of the detected vehicles from affecting the calculated velocity, a subsequent computer vision step is performed where the license plate of each individual vehicle is recognized based on a canny edge detection algorithm and a rectangular bounding box is drawn around the license plate. Repeating this for each individual frame in the video image material and observing the change of size of the license plate, the velocity of the vehicle is estimated based on the intercept theorem.
Due to increasing digitalization, the wireless reading of utility meters is gaining in importance. In this paper, a mixed integer program is presented which models a network planning problem in terms of base-station locations vis-a-vis cost, reachability and several technical limitations. The goal of the model is to find a network configuration to remotely read battery driven utility meters at optimal total costs using various low power wide-area network radio frequency technologies. It will be shown how to considerably reduce the number of variables and constraints while retaining global optimality which allows to significantly reduce the computation time required for solving instances of the model.