Search Results

Multi-sensor information fusion framework is the eyes for unmanned driving and Advanced Driver Assistance System (ADAS) to perceive the surrounding environment. In addition to the perception of the surrounding environment, real-time vehicle localization is also the key and difficult point of unmanned driving technology. The disappearance of high-precision GPS signal suddenly and defect of the lane line will bring much more difficult and dangerous for vehicle localization when the vehicle is on unmanned driving. In this paper, a road boundary feature extraction algorithm is proposed based on multi-sensor information fusion of automotive radar and vision to realize the auxiliary localization of vehicles. Firstly, we designed a 79GHz (78-81GHz) Ultra-Wide Band (UWB) millimeter-wave radar, which can obtain the point cloud information of road boundary features such as guardrail or green belt and so on.

The radar-based advanced driver assistance systems (ADAS) like autonomous emergency braking (AEB) and forward collision warning (FCW) can reduce accidents, so as to make vehicles, drivers and pedestrians safer. For active safety, automotive millimeter-wave radar is an indispensable role in the automotive environmental sensing system since it can work effectively regardless of the bad weather while the camera fails. One crucial task of the automotive radar is to detect and distinguish some objects close to each other precisely with the increasingly complex of the road condition. Nowadays almost all the automotive radar products work in bidimensional area where just the range and azimuth can be measured. However, sometimes in their field of view it is not easy for them to differentiate some objects, like the car, the manhole covers and the guide board, when they align with each other in vertical direction.

Intelligent vehicles can improve traffic safety and reduce damage caused by traffic accidents. Environmental perception system is the core of the intelligent vehicle which detects vehicles and pedestrians around the ego host-vehicle by using vehicle environmental perception sensors. Environmental perception system with the multi-sensor information fusion algorithm can utilize the advantages of each environmental perception sensor and detects targets with higher detection probability and precision. Most of the published papers are based on the sensor level fusion architecture which is not stable and robust in detecting target. This paper presents a multi-sensor fusion algorithm with central level architecture, which can improve the target detection probability compare to these with the sensor level fusion architecture.

Millimeter-wave automotive radars can prevent traffic accidents and save human lives as they can detect vehicles and pedestrians even in night and in bad weather. Various types of automotive radars operating at 24 and 77 GHz bands are developed for various applications, like adaptive cruise control, blind-spot detection and lane change assistance. In each year, millions of millimeter-wave radar are sold worldwide. Millimeter-wave radar is composed of radar hardware and radar signal processing software, which detects the targets among noise, measures the distance, longitudinal speed and the azimuth angle of the targets, tracks the targets continuously, and controls the ego vehicle to brake or accelerate. Performance of the radar signal processing software is closely related with the radar hardware properties and radar measurement conditions.

Vehicle and pedestrian detection technology is the most important part of advanced driving assistance system (ADAS) and automatic driving. The fusion of millimeter wave radar and camera is an important trend to enhance the environmental perception performance. In this paper, we propose a method of vehicle and pedestrian detection based on millimeter wave radar and camera. Moreover, the proposed method complete the detection of vehicle and pedestrian based on dynamic region generated by the radar data and sliding window. First, the radar target information is mapped to the image by means of coordinate transformation. Then by analyzing the scene, we obtain the sliding windows. Next, the sliding windows are detected by HOG features and SVM classifier in a rough detect. Then using the match function to confirm the target. Finally detecting the windows in a precision detection and merging the detecting windows. The target detection process is carried out in the following three steps.

In the last years, in order to fit the requirements of automotive radar application under the multi-target conditions, several proposals about Continuous Waveform (CW)have been developed. The transmit signal with Multiple Frequency Shift Keying (MFSK) technology was developed to analyze the target information in range domain and Doppler frequency domain simultaneously, but the MFSK waveform has lower estimation accuracy in phase measuring. A higher accuracy signal type is the chirp sequence waveform of monopulse radar, which is based on two-dimension independent frequency measuring. It can also get the range and velocity information, but might lead to ambiguities in Doppler domain. To avoid the Doppler ambiguity, a method is proposed in this paper, which uses the modified chirp sequence waveform. The carrier frequencies of the modified chirp sequence are different, which causes the Doppler frequency offset.

Millimeter wave (MMW) automotive radar plays an important role in the advanced driving assistance system (ADAS), which detects vehicles, pedestrians and other obstacles. In the adaptive cruise control (ACC) and the automatic emergency brake (AEB) system, the target needs to be oriented. One of the automotive radar’s task is to get the direction information which includes the range, speed, azimuth and height of the target by high intermediate frequency (IF) signal sampling rate. In order to solve the problem of high sampling rate for the MMW radar caused by the traditional Nyquist sampling theorem when the target is located, a new method based on the compressed sensing (CS) for the target location is proposed in this paper. This paper presents the linear frequency modulated continuous wave (LFMCW) model and simulates the sampling and reconstruction of the radar’s IF signal via CS technique by using MATLAB.

For sensing system, the trustworthiness of the variant sensors is the crucial point when dealing with advanced driving assistant system application. In this paper, an approach to a hybrid camera-radar application of vehicle tracking is presented, able to meet the requirement of such demand. Most of the time, different types of commercial sensors available nowadays specialize in different situations, such as the ability of offering a wealth of detailed information about the scene for the camera or the powerful resistance to the severe weather for the millimeter-wave (MMW) radar. The detection and tracking in different sensors are usually independent. Thus, the work here that combines the variant information provided by different sensors is indispensable and worthwhile. For the real-time requirement of merging the measurement of automotive MMW radar in high speed, this paper first proposes a fast vehicle tracking algorithm based on image perceptual hash encoding.

In autonomous driving system, lane change decision-making plays an important role as the front-end of lateral control. However, the current prediction methods of lane change are typically performed by using basic variables as the features of model without deep processing, which reduces the accuracy of the prediction. Therefore, we propose a binary logistic regression method to solve the lane change decision problem under expressway conditions, which treat quantified willingness and risk as the inputs. Firstly, we design Lane Changing Willingness function and Lane Changing Risk function with the minimum safety spacing theory and traffic environment factors. Secondly, a binary logistic regression method for predicting lane change behavior is proposed. Thirdly, we develop the driving simulation platform with low latency data collecting tools and design the experiments.

At present, the research on fatigue driving at home and abroad mainly has the following three methods: (i) driving behavioral (vehicle-based), (ii) driver behavioral (video-based), and (iii) driver physiological signals measure. The physiology-based methods have the highest recognition result. When drivers are in a state of fatigue, the Autonomic Nervous System (ANS) activity will be reflected from the physiological signal. Most of the contact sensors are used to obtain the physiological signal information of the driver. However, the contact sensors will affect the driver's driving operation, so this paper uses the frequency-modulated continuous-wave (FMCW) radar to collect the physiological signals. A fatigue driving simulation experiment was designed to collect experimental subjects' physiological signal data and separate the steady heartbeat and respiratory signals.

Lane marking detection plays a crucial role in Autonomous Driving Systems or Advanced Driving Assistance System. Vision based lane marking detection technology has been well discussed and put into practical application. LiDAR is more stable for challenging environment compared to cameras, and with the development of LiDAR technology, price and lifetime are no longer an issue. We propose a lane marking detection algorithm based on solid-state LiDARs. First a series of data pre-processing operations were done for the solid-state LiDARs with small field of view, and the needed ground points are extracted by the RANSAC method. Then, based on the OTSU method, we propose an approach for extracting lane marking points using intensity information.

Millimeter wave automotive radar is one of the most important sensors in the Advanced Driver Assistance System (ADAS) and autonomous driving system, which detects the target vehicles around the ego vehicle via processing transmitted and echo signals. However, the sampling rate of classical radar signal processing methods based on Nyquist sampling theorem is too high and the resolution of range, velocity and azimuth can’t meet the requirement of highly autonomous driving, especially azimuth. In spatial domain, targets are sparse distribution in the detection range of automotive radar. To solve these problems, the algorithm for targets location based on compressed sensing for automotive radar is proposed in this paper. Besides, the feasibility of the algorithm is verified through the simulation experiments of traffic scene. The range-doppler-azimuth model can be used to estimate the distance, velocity and azimuth of the target accurately.

The utilization of the SLAM (Simultaneous Localization and Mapping) technique was extended from the robotics to the autonomous vehicles for achieving the positioning. However, SLAM cannot obtain the global position of the vehicle but a relative one to the start. For sake of this, a fast and accurate system was proposed to obtain both the local position and the global position of vehicles based on mono-SLAM which realized the SLAM by using monocular camera with a lower cost and power consumption. Firstly, the rough latitude and longitude of current position was obtained by using common GPS without differential signal. Then, the Mono-SLAM operated on the consecutive video frames to generate the localization and local trajectory and its accuracy was further improved by utilizing the IMU information. After that, a piece of Map centered in the rough position obtained by common GPS was downloaded from the Open Street Map.

Target detection is essential to the advanced driving assistance system (ADAS) and automatic driving. And the data fusion of millimeter wave radar and camera could provide more accurate and complete information of targets and enhance the environmental perception performance. In this paper, a method of vehicle and pedestrian detection based on the data fusion of millimeter wave radar and camera is proposed to improve the target distance estimation accuracy. The first step is the targets data acquisition. A deep learning model called Single Shot MultiBox Detector (SSD) is utilized for targets detection in consecutive video frames captured by camera and further optimized for high real-time performance and accuracy. Secondly, the coordinate system of camera and radar are unified by coordinate transformation matrix. Then, the parallel Kalman filter is used to track the targets detected by radar and camera respectively.

Validation is one of the main challenges in development of automated driving systems (ADS). Due to the complexity of these systems and the various influence factors on their functional safety, current testcase generation methods can hardly guarantee the completeness and effectivity of the validation on system level. Separate validation of system components is a way to make system approval possible. In this paper, an approach is presented to generate deductively testcases for the validation of the environment perception sensors, which are the most essential components of ADS. This approach is originated from the model-based testing method, which is commonly used to validate software-based systems and extended by considering various external influence factors as follows: By modeling and analyzing applications in ADS, application oriented usecases of perception sensors are first derived.

Automotive radar is an important environment perception sensor for advance driving assistance system. It can detect objects around the vehicle with high accuracy and it works in all bad weathers. For traditional automotive radar, it cannot measure the objects’ height. Thus, a manhole cover on the road surface or a guideboard high above the road would be taken erroneously as a non-moving car. In such cases, the adaptive cruise system would decelerate or stop the vehicle erroneously and make the driver uncomfortable. A 3D automotive radar with two-dimensional electronic scanning can measure the targets’ height as well as the targets’ azimuth angle. This paper presents a 79 GHz ultra-wide band automotive 3D imaging radar. Due to the 4 GHz wide bandwidth, the range resolution of this radar can be as small as 3.75 cm.

Real-time and reliable perception of the surrounding environment is an important prerequisite for advanced driving assistance system (ADAS) and automatic driving. And vision-based detection plays a significant role in environment perception for automatic vehicles. Although deep convolutional neural networks enable efficient recognition of various objects, it has difficulty in accurately detecting special vehicles, rocks, road pile, construction site, fence and so on. In this work, we address the task of traffic scene understanding with semantic image segmentation. Both driveable area and the classification of object can be attained from the segmentation result. First, we define 29 classes of objects in traffic scenarios with different labels and modify the Deeplab V2 network. Then in order to reduce the running time, MobileNet architecture is applied to generate the feature map instead of the original models.

Estimating the motion state of peripheral targets is a very important part in the environment perception of intelligent vehicles. The accurate estimation of the motion state of the peripheral targets can provide more information for the intelligent vehicle planning module which means the intelligent vehicle is able to anticipate hazards ahead of time. To get the motion state of the target accurately, the target’s range, velocity, orientation angle and yaw rate need to be estimated. Three-dimensional high-resolution automotive millimeter wave radar can measure radial range, radial velocity, azimuth angle and elevation angle about multiple reflections of an extended target. Thus, the three-dimensional range information and three-dimensional velocity information can be obtained. With multiple three-dimensional high-resolution automotive millimeter-wave radar, it is possible to measure information in various directions of a target.

The camera-based advanced driver assistance systems (ADAS) like lane departure warning system (LDWS) and lane keeping assist (LKA) can make vehicles safer and driving easier. Lane detection is indispensable for these lane-based systems for achieving vehicle local localization and behavior prediction. Since the vision is vulnerable to the variable environment conditions such as bad weather, occlusions and illumination, the robustness is important. In this paper, a robust algorithm for detecting and tracking multiple lanes with arbitrary shape is proposed. We extend the previously lane detection and tracking process from the space domain to the temporal-spatial domain by using a more robust and general multi-lane model. First, new slice images containing temporal information are generated from image sequences. Instead of binarization process, we use a more general detector for extracting the lane marker candidates with prior knowledge to generate the binary slice image.

Automotive radar is the most important component in the autonomous driving system, which detects the obstacles, vehicles and pedestrians around with acceptable cost. The target tracking is one of the key functions in the automotive radar which estimates the position and speed of the targets having regarding to the measurement inaccuracy and interferences. Modern automotive radar requires a multi-target tracking algorithm, as in the radar field of view hundreds of targets can present. In practice, the automotive radar faces very complicated and fast-changing road conditions, for example tunnels and curved roads. The targets’ unpredictable movements and the reflections of the electromagnetic wave from the tunnel walls and the roads will make the multi-target tracking a difficult task. Such situation may last several seconds so that the continuous tracks of the targets cannot be maintained and the tracks are dropped mistakenly.