Search Results

Multiple object detection and tracking are central aspects of modeling the environment of autonomous vehicles. Lidar is a necessary component in the autonomous driving system. Without Lidar sensors, we will most probably not see fully self-driving cars become a reality. Lidar sensing gives us high-resolution data by sending out thousands of laser signals. In advanced driver assistance systems or automated driving systems, 3-D point clouds from lidar scans are typically used to measure physical surfaces. Lidar is a powerful sensor that you can use in challenging environments where other sensors might prove inadequate. Lidar can provide a complete 360-degree view of a scene. This paper designs Lidar based multi-target detection and tracking system based on the traditional point cloud processing method including down-sampling, denoising, segmentation, and clustering objects.

The recent proliferation of perception sensing and computing technologies has promoted the rapid development of automated driving. The design of the perception sensing system has nonnegligible influences both on the performances of various automated driving features and on the system costs. This paper proposes an automated driving feature oriented framework for automatic selection and arrangement of the sensors in the perception sensing system. An automated driving feature oriented optimization model is built considering the characteristics and requirements of the specific feature and a genetic algorithm based design method is provided to solve this optimization model. Furthermore, the Adaptive Cruise Control feature and the Automated Parking Assistance feature are selected as the simulation cases to verify the effectiveness of the proposed method.

To resolve the issue of poor generality of current parking methods based on idle space, a panoramic parking path planning method based on DMPR-PS and Hybrid A-Star is proposed. In this paper, a pure vision approach is adopted, and a fisheye camera is selected as an image acquisition device to build panoramic parking assistance system. Subsequently, the parking slot directional marking points are obtained through DMPR-PS model, and the valid entrance line can be also acquired to recognize parking slot. The accuracy of parking slots recognition can reach 96.86%.Eventually, the planned path can be achieved based on the Hybrid A-Star algorithm where vehicle kinematics and obstacle avoidance constraints are considered. The effectiveness of the planned path can be verified with the adoption of MATLAB and real-vehicle test.

As one of the essential parts of automatic parking system (APS), the parking motion control module directly affects the system performance and driver experience. Therefore, it is necessary to design a simple, robust and efficient trajectory tracking algorithm which adapt to the various parking conditions. Firstly, considering the predictability and the ability of dealing with various system constraints, the model predictive control (MPC) lateral controller is designed. Then, the second lateral controller based on linear quadratic regulator (LQR) algorithm is designed, which has the excellent capability of balancing the multiple performances of the system. Finally, Stanley lateral controller is designed as the benchmark for horizontal comparison. Parallel and vertical parking simulation environments are proposed to verify the effectiveness of the designed lateral controllers, and the advantages and shortcomings of each control algorithm are horizontally analyzed and summarized.

In the V2X system, the correct decisions in many scenarios require the V2X device node clock synchronization, so that the time or frequency error between each network node device remains within a reasonable error level. The high-precision time service of multi-source fusion is more important in the scene where GNSS signal is blocked or multi-sensor fusion requires high signal accuracy. In addition, in the automatic driving technology, the time delay generated at different stages of perception, planning and execution needs to be considered. The signal positioning of running vehicles needs to control the synchronous time error at the nanosecond level. This paper provides a high-precision time synchronization method applied to V2X. The method adopts multi clock source fusion. It includes three service modes: GNSS, cellular data and local clock.

In the fusion-based vehicle positioning, the lane detection is applied to provide the relative position of ego-vehicle and lanes, which is critical to subsequent tasks including trajectory planning and behaviour decision. However, the performance of current vision-based lane detectors drop significantly when facing adversarial visual conditions, e.g., low-illumination conditions like night scenarios. Images captured in this scenario often suffer from low contrast, low brightness and noise, which is challenging for detectors to extract correct information. To facilitate the lane detection in low-illumination conditions, this paper presents a novel framework which integrates image feature enhancement with lane detection. The framework consists of two modules: an image enhancement module to enhance and extract information from low visibility images, and a detection module to regress the lane parameters. Both modules are optimized by loss collaboration.

Multi-Target tracking is a central aspect of modeling the surrounding environment of autonomous vehicles. Automotive millimeter-wave radar is a necessary component in the autonomous driving system. One of the biggest advantages of radar is it measures the velocity directly. Another big advantage is that the radar is less influenced by environmental conditions. It can work day and night, in rainy or snowy conditions. In the expressway scenario, the forward-looking radar can generate multiple objects, to properly track the leading vehicle or neighbor-lane vehicle, a multi-target tracking algorithm is required. How to associate the track and the measurement or data association is an important question in a multi-target tracking system. This paper applies the nearest-neighbor method to solve the data association problem and uses an extended Kalman filter to update the state of the track.

Aiming at the problems of ineffective collision avoidance and vehicle instability in the process of vehicle emergency braking in road conditions with low adhesion and sudden change in adhesion coefficient, a stability-coordinated emergency braking and collision avoidance control system SEBCACS) is proposed. First, according to the motion of the ego vehicle and the target vehicle as well as the road adhesion conditions, a collision time model is proposed for evaluating the vehicle collision risk, and the expected deceleration required to avoid the collision is calculated. Then, the MPC method is used to calculate the yaw moment generated by the four-wheel braking force required to maintain vehicle stability according to the actual and reference yaw rate and side slip angle deviation. Then it is decided whether to implement additional yaw moment control according to the body stability evaluation results.

The Autonomous emergency braking system (AEB) has been widely equipped in the design and manufacture of vehicles as an active safety system for preventing rear-end collisions. It has shown great safety potential in preventing collisions and reducing collision injuries. However, there are differences in the response characteristics of drivers in emergency scenarios due to individual differences and driving habits. The impact of different driver types on the safety performance of AEB systems has not been evaluated. In this study, the typical driver response model was constructed by selecting driver response features representing alertness and braking. The AEB algorithm of distance and situation awareness was combined with the kinematic of vehicle before the collision to construct a simulation case based on the rear-end collisions in the China in-depth accident study database (CIDAS).

An autonomous vehicle is able to perceive and interpret exactly its surroundings and its interior (“Sensing”). then, it processes the information received and plan its driving strategy (“processing”). And finally, it uses its powertrain, steering and braking power to move its wheels in such a way that the planned driving strategy is put into practice (“Acting”). Testing an autonomous vehicle’s reaction to the erratic traffic scenarios using prototypes would be impractical. Physically testing these scenarios can also be risky to human life and equipment. Additionally, the repetition involved in the comprehensive testing of all these scenarios could lead to human errors. Various Self Driving car manufacturers have reported injuries and causalities while doing Functional testing [1].

Simulation of real time situations is a time tested software validation methodology in the automotive industry and array of simulation technologies have been in use for decades and is widely accepted and been part & parcel of software development cycle. While software that is being developed needs detailed plan, architecture and detailed design, it also matters during its development that, it is built in the right way from the very beginning and is fine tuned constantly. Especially for Software-In-Loop simulation (SIL), plenty of practices/tools/techniques/data are being used for simulation of system/software behavior. When it comes to choosing the right simulation technique and tools to be adopted, often there are discussions revolve around cost, feasibility, effectiveness, man-power, scalability, reusability etc.

Advanced Driver Assistance Systems (ADAS) rely on camera sensors to work effectively to provide warning signs to prevent forward and rearward collision, lane departure, pedestrian detection, traffic sign recognition, automatic headlight control and assists for autonomous driving. Generally, vision based ADAS systems have wide angle cameras installed on front, rear and sides of the vehicle. These camera-based vision sensors are subjected to severe thermal environments that can impact its sensing performance and image quality. Hence it is imperative to thermally qualify the camera module to ensure reliable performance without loss in functionality. The thermal environment experienced by these cameras vary based on their mounting location. Intelligent forward view cameras are mounted in windshield region of the vehicle and encounter sun load in order of 1000 W/m2.

Vehicle manufacturers conduct tests to develop crash sensing system calibrations. Ditch fall-over is one of a suite of laboratory tests used to develop rollover sensing calibrations that can trigger deployment of safety devices like roof rail airbags and seat belt pretensioners. The ditch fall-over test simulates a flat road followed by a ditch on one side of the road. The vehicle heads into the ditch and the driver applies swift steer input once the ditch slope is sensed. Typically, the steer input is applied when the two down-slope wheels on the ditch side enter the ditch. Multi-Body Dynamics (MBD) software can be used for virtual simulation of these test events. Conventionally in simulations, the vehicle-model is run without steer input and the marking line crossing time is observed/manually recorded from observation of simulation video. This recorded time is used to apply the steer input and the full event is then re-simulated.

77 GHz band FMCW radar is used, which receives a non-stationary sequence of Radar Cross-Section (RCS) for every detected object. These RCS schemes are used in automotive radar systems. Most established non-stationary sequence classification approaches use radar images or a complex signal modeling method. Due to less on-chip memory, stringent run-time requirements, and space complexity of the problem, we proposed a novel way of extracting representative features to avoid false braking events with a focus on memory and code optimization. This paper deals with a feature-based sequence classification problem wherein features are extracted by identifying patterns and trends, which are then used by the machine learning classification model. In this study, we take advantage of linear fitting, curve fitting, exploratory plot analysis, and statistical analysis to create distinguishing features. At every radar cycle, long-range radar can detect a certain number of objects from its environment.

Autonomous vehicles (AVs) are self-contained vehicles equipped with control systems to execute various tasks. The Lane Departure Warning (LDW) system is widely employed to prevent the most common cause of vehicle collisions. An autonomous lane-departure system will aid and reduce such collisions. When the vehicle is at risk of drifting or departing its lane, the LDW system monitors its relative position to the lane edge and sends an alarming warning signal to the driver. This work uses an ML-based technique to detect lane markers in an Indian context using a high-resolution camera mounted on the car. Considering that, the LDW system requires three primary operations. The camera geometry information is used to divide the acquired image into two parts: a road part and a non-road component. Then, to circumvent the obstacles caused by the perspective effect, inverse perspective mapping is applied.

Automotive industry is going through a massive digital transformation to enable advance ADAS functions like cruise control, safety and parking assist. To develop and test advance and complex deep neural network-based AI/ML ADAS models, the need of huge amount of rich and diverse annotated data is utmost important. Over the past decade it has been observed that annotation complexity has increased tremendously and evolved from a simple bounding box to complex annotations like segmentation, 3D bounding box, key points etc. that too with multiple sensor integration. Hence such stupendous annotation task cannot be executed inhouse unlike in the past, companies choose to outsource time consuming and labor-intensive task to third party vendors. Hence annotation becomes an additional and unexpected challenge in ADAS function development, which urge the need for standard annotation format.

Pothole detection in Intelligent Transportation Systems (ITS) vehicles has been a part of Advanced driver-assistance systems (ADAS) for a long time. Various sensors have been used for this purpose so far: Accelerometer, Gyroscope, etc. However, the fusion of multiple modalities of information from different sensors remains a challenge, mainly owing to the different sampling rates and varying frame rates used by each sensor. Other sensor types like Radar and LIDAR, though precise, are difficult to use, thus forcing us to look for low-cost solutions. Our proposed work uses Accelerometer and Gyroscope sensor fusion to predict pothole presence in Indian scenarios. Previous works have mainly dealt with predicting potholes with data collected using either traditional machine learning techniques like Decision trees, (Support Vector Machines (SVM)’s and Light Gradient Boosting Machine (LGBM) and deep learning methods using neural networks and attention mechanisms.

Automobile sector is growing every day with fast affinity towards Autonomous vehicles. The most challenging task of ADAS based driverless car is to identify and track the objects in front of the vehicle. To implement this type of technology we require a robust algorithm which can classify the object just-in-time and have great accuracy. We are using automotive radar sensor of 77GHz frequency. Quite often we’ve noticed sudden fluctuations in prediction of the obstacles using either heuristic or even machine learning techniques which focus only on frame-wise / cycle-wise data. So, this inspires us to investigate the history of the data coming in as opposed to only one cycle at a time. Hence, we incorporated a technique wherein we could make use of the past data as well as current cycle data. In this paper, we’ve used Radar time series data to classify the object in front of the Ego vehicle in each Radar cycle.

Advance Driver Assistance System(ADAS) is one of the growing technologies in the automotive industry owing to the expected future demands for autonomous/driverless vehicles. The rising awareness of the driver’s safety and comfort, influence of regulations and safety ratings on Original Equipment Manufacturers (OEM) are the major factors which are responsible for enhancing the growth of the market of ADAS-equipped vehicles. Before introducing the ADAS features into the market, it is important to validate feature on vehicle in real time. To validate ADAS systems in vehicle it requires complex equipment’s and infrastructure. This includes the equipment which calculate relative position between multiple vehicles and objects in a test track under real time. Along with this vehicle test equipment’s, it also requires different soft targets with a guided moving platform. This paper presents a detailed study on different infrastructures, vehicle test equipment’s and soft targets.

With the advent of autonomous driving into mobility industry, passenger safety is of paramount importance. Electronic Stability Control (ESC) has been widely employed as an active safety feature in most of the modern cars, ESC is significantly based on differential braking, steer by wire or active torque distribution to prevent the vehicle from going off-course. However, ESC is limited to the discretion of the driver. Thus, the objective of this paper is to improve functionalities of ESC by adding sensor detection-based algorithm for collision avoidance. The subject vehicle is installed with an existing default ESC & ABS system along with two range tracking sensors at front and rear and two blind spot detection sensors. The algorithm for collision avoidance has been tested in CarSim 2021.1.