scholarly journals Processing, assessing, and enhancing the Waymo autonomous vehicle open dataset for driving behavior research

2022 ◽  
Vol 134 ◽  
pp. 103490
Author(s):  
Xiangwang Hu ◽  
Zuduo Zheng ◽  
Danjue Chen ◽  
Xi Zhang ◽  
Jian Sun
Keyword(s):  
Autonomous Vehicle ◽  
Driving Behavior

scholarly journals Improving the Learning of Self-driving Vehicles Based on Real Driving Behavior using Deep Neural Network Techniques

2020 ◽  
Author(s):  
Nayere Zaghari ◽  
Mahmood Fathy ◽  
Seyed Mahdi Jameii ◽  
Mohammad Sabokrou ◽  
Mohammad Shahverdy
Keyword(s):  
Neural Network ◽  
Deep Neural Network ◽  
Autonomous Vehicle ◽  
Driving Behavior ◽  
Obstacle Detection ◽  
Convolutional Network ◽  
The Real ◽  
Cloning Method ◽  
Vehicle Acceleration ◽  
Performance Results

Considering the significant advancements in autonomous vehicle technology, research in this field is of interest to researchers. To drive vehicles autonomously, controlling steer angle, gas hatch, and brakes need to be learned. The behavioral cloning method is used to imitate humans’ driving behavior. We created a dataset of driving in different routes and conditions and using the designed model, the output used for controlling the vehicle is obtained. In this paper, the Learning of Self-driving Vehicles Based on Real Driving Behavior Using Deep Neural Network Techniques (LSV-DNN) is proposed. We designed a convolutional network which uses the real driving data obtained through the vehicle’s camera and computer. The response of the driver is during driving is recorded in different situations and by converting the real driver’s driving video to images and transferring the data to an excel file, obstacle detection is carried out with the best accuracy and speed using the Yolo algorithm version 3. This way, the network learns the response of the driver to obstacles in different locations and the network is trained with the Yolo algorithm version 3 and the output of obstacle detection. Then, it outputs the steer angle and amount of brake, gas, and vehicle acceleration. The LSV-DNN is evaluated here via extensive simulations carried out in Python and TensorFlow environment. We evaluated the network error using the loss function. By comparing other methods which were conducted on the simulator’s data, we obtained good performance results for the designed network on the data from KITTI benchmark, the data collected using a private vehicle, and the data we collected.

Toward More Realistic Driving Behavior Models for Autonomous Vehicles in Driving Simulators

2003 ◽  
Vol 1843 (1) ◽  
pp. 41-49 ◽  
Author(s):  
Talal Al-Shihabi ◽  
Ronald R. Mourant
Keyword(s):  
Autonomous Vehicles ◽  
Driving Simulator ◽  
Autonomous Vehicle ◽  
Three Dimensional ◽  
Driving Behavior ◽  
Behavioral Models ◽  
Aggressive Driving ◽  
Decision Making Unit ◽  
Driving Model ◽  
Dynamics Models

Autonomous vehicles are perhaps the most encountered element in a driving simulator. Their effect on the realism of the simulator is critical. For autonomous vehicles to contribute positively to the realism of the hosting driving simulator, they need to have a realistic appearance and, possibly more importantly, realistic behavior. Addressed is the problem of modeling realistic and humanlike behaviors on simulated highway systems by developing an abstract framework that captures the details of human driving at the microscopic level. This framework consists of four units that together define and specify the elements needed for a concrete humanlike driving model to be implemented within a driving simulator. These units are the perception unit, the emotions unit, the decision-making unit, and the decision-implementation unit. Realistic models of humanlike driving behavior can be built by implementing the specifications set by the driving framework. Four humanlike driving models have been implemented on the basis of the driving framework: ( a) a generic normal driving model, ( b) an aggressive driving model, ( c) an alcoholic driving model, and ( d) an elderly driving model. These driving models provide experiment designers with a powerful tool for generating complex traffic scenarios in their experiments. These behavioral models were incorporated along with three-dimensional visual models and vehicle dynamics models into one entity, which is the autonomous vehicle. Subjects perceived the autonomous vehicles with the described behavioral models as having a positive effect on the realism of the driving simulator. The erratic driving models were identified correctly by the subjects in most cases.

Improving the Learning of Self-driving Vehicles Based on Real Driving Behavior using Deep Neural Network Techniques

2020 ◽  
Author(s):  
Nayere Zaghari ◽  
Mahmood Fathy ◽  
Seyed Mahdi Jameii ◽  
Mohammad Sabokrou ◽  
Mohammad Shahverdy
Keyword(s):  
Neural Network ◽  
Deep Neural Network ◽  
Autonomous Vehicle ◽  
Driving Behavior ◽  
Obstacle Detection ◽  
Convolutional Network ◽  
The Real ◽  
Cloning Method ◽  
Vehicle Acceleration ◽  
Performance Results

Considering the significant advancements in autonomous vehicle technology, research in this field is of interest to researchers. To drive vehicles autonomously, controlling steer angle, gas hatch, and brakes need to be learned. The behavioral cloning method is used to imitate humans’ driving behavior. We created a dataset of driving in different routes and conditions and using the designed model, the output used for controlling the vehicle is obtained. In this paper, the Learning of Self-driving Vehicles Based on Real Driving Behavior Using Deep Neural Network Techniques (LSV-DNN) is proposed. We designed a convolutional network which uses the real driving data obtained through the vehicle’s camera and computer. The response of the driver is during driving is recorded in different situations and by converting the real driver’s driving video to images and transferring the data to an excel file, obstacle detection is carried out with the best accuracy and speed using the Yolo algorithm version 3. This way, the network learns the response of the driver to obstacles in different locations and the network is trained with the Yolo algorithm version 3 and the output of obstacle detection. Then, it outputs the steer angle and amount of brake, gas, and vehicle acceleration. The LSV-DNN is evaluated here via extensive simulations carried out in Python and TensorFlow environment. We evaluated the network error using the loss function. By comparing other methods which were conducted on the simulator’s data, we obtained good performance results for the designed network on the data from KITTI benchmark, the data collected using a private vehicle, and the data we collected.

scholarly journals Common Driving Notification Protocol Based on Classified Driving Behavior for Cooperation Intelligent Autonomous Vehicle Using Vehicular Ad-Hoc Network Technology

2015 ◽  
Vol 5 (1) ◽  
pp. 5-21 ◽  
Author(s):  
Che-Hung Lin ◽  
Fang-Yan Dong ◽  
Kaoru Hirota
Keyword(s):  
Autonomous Vehicles ◽  
Ad Hoc ◽  
Autonomous Vehicle ◽  
Driving Behavior ◽  
Packet Transmission ◽  
Wireless Data ◽  
Standard Format ◽  
Simulation Tools ◽  
Identification Time ◽  
Basic Protocol

Abstract A protocol, called common driving notification protocol (CDNP), is proposed based on the classified driving behavior for intelligent autonomous vehicles, and it defines a standard with common messages and format for vehicles. The common standard format and definitions of CDNP packet make the autonomous vehicles have a common language to exchange more detail driving decision information of various driving situations, decrease the identification time for one vehicle to identify the driving decisions of other vehicles before or after those driving decisions are performed. The simulation tools, including NS- 3 and SUMO, are used to simulate the wireless data packet transmission and the vehicle mobility; the experiment results present that the proposed protocol, CDNP, can increase the reaction preparing time with maximum value 250 seconds, decrease the identification time and the average travel time. Prospectively, it is decided to implement the CDNP as a protocol stack in the Linux kernel to provide the basic protocol capability for real world transmission testing.

scholarly journals An Investigation of Autonomous Vehicle Roundabout Situation

2019 ◽  
Vol 48 (3) ◽  
pp. 236-241
Author(s):  
Hang Cao ◽  
Máté Zöldy
Keyword(s):  
Fuel Consumption ◽  
Autonomous Vehicles ◽  
Autonomous Vehicle ◽  
Preliminary Test ◽  
Driving Behavior ◽  
Physical Parameters ◽  
Test Scenario ◽  
Vehicle Type ◽  
Fuel Consumption And Emissions ◽  
The Impact

The aim of this paper is to evaluate the impact of connected autonomous behavior in real vehicles on vehicle fuel consumption and emission reductions. Authors provide a preliminary theoretical summary to assess the driving conditions of autonomous vehicles in roundabout, which attempts exploring the impact of driving behavior patterns on fuel consumption and emissions, and including other key factors of autonomous vehicles to reduce fuel consumption and emissions. After summarizing, driving behavior, effective in-vehicle systems, both roundabout physical parameters and vehicle type are all play an important role in energy using. ZalaZONE’s roundabout is selected for preliminary test scenario establishment, which lays a design foundation for further in-depth testing.

scholarly journals Improving the Learning of Self-driving Vehicles Based on Real Driving Behavior using Deep Neural Network Techniques

2020 ◽  
Author(s):  
Nayereh Zaghari ◽  
Mahmood Fathy ◽  
Seyed Mahdi Jameii ◽  
Mohammad Sabokrou ◽  
Mohammad Shahverdy
Keyword(s):  
Neural Network ◽  
Deep Neural Network ◽  
Autonomous Vehicle ◽  
Driving Behavior ◽  
Obstacle Detection ◽  
Convolutional Network ◽  
The Real ◽  
Cloning Method ◽  
Vehicle Acceleration ◽  
Performance Results

Considering the significant advancements in autonomous vehicle technology, research in this field is of interest to researchers. To drive vehicles autonomously, controlling steer angle, gas hatch, and brakes need to be learned. The behavioral cloning method is used to imitate humans’ driving behavior. We created a dataset of driving in different routes and conditions and using the designed model, the output used for controlling the vehicle is obtained. In this paper, the Learning of Self-driving Vehicles Based on Real Driving Behavior Using Deep Neural Network Techniques (LSV-DNN) is proposed. We designed a convolutional network which uses the real driving data obtained through the vehicle’s camera and computer. The response of the driver is during driving is recorded in different situations and by converting the real driver’s driving video to images and transferring the data to an excel file, obstacle detection is carried out with the best accuracy and speed using the Yolo algorithm version 3. This way, the network learns the response of the driver to obstacles in different locations and the network is trained with the Yolo algorithm version 3 and the output of obstacle detection. Then, it outputs the steer angle and amount of brake, gas, and vehicle acceleration. The LSV-DNN is evaluated here via extensive simulations carried out in Python and TensorFlow environment. We evaluated the network error using the loss function. By comparing other methods which were conducted on the simulator’s data, we obtained good performance results for the designed network on the data from KITTI benchmark, the data collected using a private vehicle, and the data we collected.

Driving-behavior-oriented trajectory planning for autonomous vehicle driving on urban structural road

2020 ◽  
pp. 095440702096999
Author(s):  
Dequan Zeng ◽  
Zhuoping Yu ◽  
Lu Xiong ◽  
Junqiao Zhao ◽  
Peizhi Zhang ◽  
...  
Keyword(s):  
Path Planning ◽  
Velocity Profile ◽  
Success Rate ◽  
Trajectory Planning ◽  
Autonomous Vehicle ◽  
Driving Behavior ◽  
Vehicle Stability ◽  
Lane Change ◽  
Actuator Constraints ◽  
Traffic Rules

A novel driving-behavior-oriented method is proposed in this paper for improving trajectory planning performance of autonomous vehicle driving on urban structural road. Differ from the irregularity and unpredictability of escaping a maze or travelling on off-road, the driving on road emphasizes more on the compliance of road traffic rules and the satisfaction of passenger comfort rather than purely pursuing the shortest route or the shortest time. Therefore, the driving-behavior-oriented framework is employed in trajectory planning, which divides trajectory into lane change, turn and U-turn, according to the basic traffic rules and the daily behaviors of drivers. The presented approach mainly includes basic path planning, fast-bias RRT path planning and velocity planning. The basic path planning consists of lane change, turn and U-turn behaviors, which generates smooth path with continuous curvature. In order to ensure the completeness of the programming algorithm, a fast-bias RRT (FB-RRT) algorithm is embedded. As guiding by the driving behavior, normal random, goal-bias and Gaussian sampling strategies are fused to form FB-RRT, which could make the best use of the basic path planning and reduce the randomness of node’s extension to save the computation time. After collision-free path generating, cubic polynomial curve is employed to schedule velocity profile for coping with vehicle stability requirements, actuator constraints and comfort conditions. The planner has been tested in simulation and a real vehicle in various typical scenarios. Test results illustrate that the presented method could generate a trajectory with controllable extrema of curvature as well as with continuous and smooth enough curvature. Besides, generated trajectory has short length, high success rate (no less than 80% average success rate in complex environment) and real time (the average period is less than 100 ms). Moreover, the velocity profile meets the vehicle stability requirements, actuator constraints, and comfort conditions.

Sign in / Sign up

Export Citation Format

Share Document