For autonomous driving, it is imperative to perform various high-computation image recognition tasks with high accuracy, utilizing diverse sensors to perceive the surrounding environment. Specifically, cameras are used to perform lane detection, object detection, and segmentation, and, in the absence of lidar, tasks extend to inferring 3D information through depth estimation, 3D object detection, 3D reconstruction, and SLAM. However, accurately processing all these image recognition operations in real-time for autonomous driving under constrained hardware conditions is practically unfeasible. In this study, considering the characteristics of image recognition tasks performed by these sensors and the given hardware conditions, we investigated MTL (multi-task learning), which enables parallel execution of various image recognition tasks to maximize their processing speed, accuracy, and memory efficiency. Particularly, this study analyzes the combinations of image recognition tasks for autonomous driving and proposes the MDO (multi-task decision and optimization) algorithm, consisting of three steps, as a means for optimization. In the initial step, a MTS (multi-task set) is selected to minimize overall latency while meeting minimum accuracy requirements. Subsequently, additional training of the shared backbone and individual subnets is conducted to enhance accuracy with the predefined MTS. Finally, both the shared backbone and each subnet undergo compression while maintaining the already secured accuracy and latency performance. The experimental results indicate that integrated accuracy performance is critically important in the configuration and optimization of MTL, and this integrated accuracy is determined by the ITC (inter-task correlation). The MDO algorithm was designed to consider these characteristics and construct multi-task sets with tasks that exhibit high ITC. Furthermore, the implementation of the proposed MDO algorithm, coupled with additional SSL (semi-supervised learning) based training, resulted in a significant performance enhancement. This advancement manifested as approximately a 12% increase in object detection mAP performance, a 15% improvement in lane detection accuracy, and a 27% reduction in latency, surpassing the results of previous three-task learning techniques like YOLOP and HybridNet. Full article

Ultrasonic sensors are inexpensive and provide highly accurate measurements, even with simple hardware configurations, facilitating their use in various fields. When multiple ultrasonic sensors exist in the measurement space, crosstalk occurs due to other nodes, which leads to incorrect measurements. Crosstalk includes not only receiving homogeneous signals from other nodes, but also overlapping by other signals and interference by heterogeneous signals. This paper proposes using frequency sweep keying modulation to provide robustness against overlap and a faster region-based convolutional neural network (R-CNN) demodulator to reduce the interference caused by heterogeneous signals. The demodulator works by training Faster R-CNN with the spectrograms of various received signals and classifying the received signals using Faster R-CNN. Experiments implementing an ultrasonic crosstalk environment showed that, compared to on–off keying (OOK), phase-shift keying (PSK), and frequency-shift keying (FSK), the proposed method can implement CDMA even with shorter codes and is robust against overlap. Compared to correlation-based frequency sweep keying, the time-of-flight error was reduced by approximately 75%. While the existing demodulators did not consider heterogeneous signals, the proposed method ignored approximately 99% of the OOK and PSK signals and approximately 79% of the FSK signals. The proposed method performed better than the existing methods and is expected to be used in various applications. Full article

This paper proposes a novel automotive radar waveform involving the theory behind M-ary frequency shift key (MFSK) radar systems. Along with the MFSK theory, coding schemes are studied to provide a solution to mutual interference. The proposed MFSK waveform consists of frequency increments throughout the range of 76 GHz to 81 GHz with a step value of 1 GHz. Instead of stepping with a fixed frequency, a triangular chirp sequence allows for static and moving objects to be detected. Therefore, automotive radars will improve Doppler estimation and simultaneous range of various targets. In this paper, a binary coding scheme and a combined transform coding scheme used for radar waveform correlation are evaluated in order to provide unique signals. AVs have to perform in an environment with a high number of signals being sent through the automotive radar frequency band. Efficient coding methods are required to increase the number of signals that are generated. An evaluation method and experimental data of modulated frequencies as well as a comparison with other frequency method systems are presented. Full article

This paper describes the conception of a high level, compact, scalable, and long autonomy perception and localization system for autonomous driving applications. Our benchmark is composed of a high resolution lidar (128 channels), a stereo global shutter camera, an inertial navigation system, a time server, and an embedded computer. In addition, in order to acquire data and build multi-modal datasets, this system embeds two perception algorithms (RBNN detection, DCNN detection) and one localization algorithm (lidar-based localization) to provide real-time advanced information such as object detection and localization in challenging environments (lack of GPS). In order to train and evaluate the perception algorithms, a dataset is built from 10,000 annotated lidar frames from various drives carried out under different weather conditions and different traffic and population densities. The performances of the three algorithms are competitive with the state-of-the-art. Moreover, the processing time of these algorithms are compatible with real-time autonomous driving applications. By providing directly accurate advanced outputs, this system might significantly facilitate the work of researchers and engineers with respect to planning and control modules. Thus, this study intends to contribute to democratizing access to autonomous vehicle research platforms. Full article

Vehicle dynamics control systems have a fundamental role in smart and autonomous mobility, where one of the most crucial aspects is the vehicle body velocity estimation. In this paper, the problem of a correct evaluation of the vehicle longitudinal velocity for dynamic control applications is approached using a neural networks technique employing a set of measured samples referring to signals usually available on-board, such as longitudinal and lateral acceleration, steering angle, yaw rate and linear wheel speed. Experiments were run on four professional driving circuits with very different characteristics, and the vehicle longitudinal velocity was estimated with different neural network training policies and validated through comparison with the measurements of the one acquired at the vehicle’s center of gravity, provided by an optical Correvit sensor, which serves as the reference (and, therefore, exact) velocity values. The results obtained with the proposed methodology are in good agreement with the reference values in almost all tested conditions, covering both the linear and the nonlinear behavior of the car, proving that artificial neural networks can be efficiently employed onboard, thereby enriching the standard set of control and safety-related electronics. Full article

Currently, research on automated vehicles is strongly related to technological advances to achieve a safe, more comfortable driving process in different circumstances. The main achievements are focused mainly on highway and interurban scenarios. The urban environment remains a complex scenario due to the number of decisions to be made in a restrictive context. In this context, one of the main challenges is the automation of the relocation process of car-sharing in urban areas, where the management of the platooning and automatic parking and de-parking maneuvers needs a solution from the decision point of view. In this work, a novel behavioral planner framework based on a Finite State Machine (FSM) is proposed for car-sharing applications in urban environments. The approach considers four basic maneuvers: platoon following, parking, de-parking, and platoon joining. In addition, a basic V2V communication protocol is proposed to manage the platoon. Maneuver execution is achieved by implementing both classical (i.e., PID) and Model-based Predictive Control (i.e., MPC) for the longitudinal and lateral control problems. The proposed behavioral planner was implemented in an urban scenario with several vehicles using the Carla Simulator, demonstrating that the proposed planner can be helpful to solve the car-sharing fleet relocation problem in cities. Full article

Various sensors are embedded in automobiles to implement intelligent safety technologies such as autonomous driving and front–rear collision avoidance technology. In particular, ultrasonic sensors have been used in the past because they have an accuracy of centimeters to sub-centimeters in air despite their low cost and low hardware complexity. Recently, the crosstalk problem between ultrasonic sensors has been raised because the number of ultrasonic sensors in the unit space has increased as the number of vehicles increases. Various studies have been conducted to solve the crosstalk, but a demodulation error occurs when signals overlap. Therefore, in this paper, we propose a method that is robust to ultrasonic signal overlap, is robust even at shorter code length, and has reduced time of flight (TOF) error compared to the existing method by applying frequency sweep keying modulation based on code division multiple access (CDMA). As a result of the experiment, the code was detected accurately regardless of the overlap ratio of the two signals, and it was robust even in situations where the power of the two signals was different. In addition, it shows an accurate TOF estimation even if the ID code length is shorter than the existing on–off-keying, frequency shift keying, and phase shift keying methods. Full article

The introduction of autonomous vehicles has been considered as a possible option for reducing traffic congestion in many transport studies. Many types of models, in particular car-following microsimulation models have been adopted in most studies. The impacts of autonomous vehicles (AVs) on congestion, however, have not yet been concluded. This could be because different researchers use different forms of car-following models to assess these impacts, or because the utilised modelling approaches and their parameters are different in different studies. In particular, two of the important parameters that are associated with car-following models are the used values for maximum acceleration and the average desired time gaps. While the values of these parameters can be adjusted and controlled by the ACC controllers in the AV, they can also be controlled by the users. Therefore, assigning unrealistic values to these parameters could well result in unrealistic conclusions. This paper investigated the impacts of the maximum acceleration and the average desired time gaps on congestion levels using the loss-time indicator. The analysis was carried out on the Hanshin expressway in Japan and was tested and assessed using the Helly (FACC) car-following microsimulation model. This includes estimating the values of the desired time gap from real traffic time-gap distributions. The Hanshin expressway is an urban toll highway of 273 km that extends from Osaka to Kobe, representing the Hanshin area in Japan. The Hanshin highway serves a huge traffic volume that consists of private and freight vehicles that operate within the Hanshin area. This area represents one of three major municipal areas in Japan including Tokyo and Nagoya. A total of 740,000 vehicles per day travel on the expressway. As a result, there is significant congestion on the Hanshin expressway. There have been various plans put in place to ease congestion ranging from building new roads to the implementation of traffic-demand-management measures. However, the predictions of the impacts of such measures do not provide any evidence that they would ease traffic congestion. Other possible measures that could be investigated for easing traffic congestion include technology-based solutions such as autonomous vehicles. The modelling results recommend that the results obtained from microsimulation models should be taken with care, and good attention should be paid to the parameters used and their values in the model. The values assigned to driving-behaviour parameters, the maximum values of acceleration, and the time-gap settings, for example, control the final outcomes of the models. Full article

When it comes to some essential abilities of autonomous ground vehicles (AGV), detection is one of them. In order to safely navigate through any known or unknown environment, AGV must be able to detect important elements on the path. Detection is applicable both on-road and off-road, but they are much different in each environment. The key elements of any environment that AGV must identify are the drivable pathway and whether there are any obstacles around it. Many works have been published focusing on different detection components in various ways. In this paper, a survey of the most recent advancements in AGV detection methods that are intended specifically for the off-road environment has been presented. For this, we divided the literature into three major groups: drivable ground and positive and negative obstacles. Each detection portion has been further divided into multiple categories based on the technology used, for example, single sensor-based, multiple sensor-based, and how the data has been analyzed. Furthermore, it has added critical findings in detection technology, challenges associated with detection and off-road environment, and possible future directions. Authors believe this work will help the reader in finding literature who are doing similar works. Full article