Search Results

Intelligent driving, aimed for collision avoidance and self-navigation, is mainly based on environmental sensing via radar, lidar and/or camera. While each of the sensors has its own unique pros and cons, camera is especially good at object detection, recognition and tracking. However, unpredictable environmental illumination can potentially cause misdetection or false detection. To investigate the influence of illumination conditions on detection algorithms, we reproduced various illumination intensities in a photo-realistic virtual world, which leverages recent progress in computer graphics, and verified vehicle detection effect there. In the virtual world, the environmental illumination is controlled precisely from low to high to simulate different illumination conditions in the driving scenarios (with relative luminous intensity from 0.01 to 400). Sedan cars with different colors are modelled in the virtual world and used for detection task.

Adaptive cruise control system with lane change assistance (LCACC) is a novel advanced driver assistance system (ADAS), which enables dual-target tracking, safe lane change, and longitudinal ride comfort. To design the personalized LCACC system, one of the most important prerequisites is to identify the driver’s individualities. This paper presents a real-time driver behavior characteristics identification strategy for LCACC system. Firstly, a driver behavior data acquisition system was established based on the driver-in-the-loop simulator, and the behavior data of different types of drivers were collected under the typical test condition. Then, the driver behavior characteristics factor Ks we proposed, which combined the longitudinal and lateral control behaviors, was used to identify the driver behavior characteristics. And an individual safe inter-vehicle distances field (ISIDF) was established according to the identification results.

Due to growing interest in automated driving, the need for better understanding of human driving behavior in uncertain environment, such as driving behavior at un-signalized crossroad and roundabout, has further increased. Driving behavior at roundabout is greatly influenced by different dynamic factors such as speed, distance and circulating flow of the potentially conflicting vehicles, and drivers should choose whether to leave or wait at the upcoming exit according to these factors. In this paper, the influential dynamic factors and driving behavior characteristics at the roundabout is analyzed in detail, random forest method is then deployed to predict the driving behavior. For training the driving behavior model, four typical roundabout layouts were created under a real-time driving simulator with PanoSim-RT and dSPACE. Traffic participants with different motion style were also set in the simulation platform to mimic real driving conditions.

Driver individualities is crucial for the development of the Advanced Driver Assistant System (ADAS). Due to the mechanism that specific driving operation action of individual driver under typical conditions is convergent and differentiated, a novel driver individualities recognition method is constructed in this paper using random forest model. A driver behavior data acquisition system was built using dSPACE real-time simulation platform. Based on that, the driving data of the tested drivers were collected in real time. Then, we extracted main driving data by principal component analysis method. The fuzzy clustering analysis was carried out on the main driving data, and the fuzzy matrix was constructed according to the intrinsic attribute of the driving data. The drivers’ driving data were divided into multiple clusters.

It’s hard to achieve complete self-driving using hand-crafting generalized decision-making rules, while the end-to-end self-driving system is low in complexity, does not require hand-crafting rules, and can deal with complex situations. Modular-based self-driving systems require multi-task fusion and high-precision maps, resulting in high system complexity and increased costs. In end-to-end self-driving, we usually only use camera to obtain scene status information, so image processing is very important. Numerous deep learning applications benefit from multi-task learning, as the multi-task learning can accelerate model training and improve accuracy with combine all tasks into one model, which reduces the amount of calculation and allows these systems to run in real-time. Therefore, the approach of obtaining rich scene state information based on multi-task learning is very attractive. In this paper, we propose an approach to multi-task learning for semantics, geometry and motion.

Most of the Advanced Driver Assistance System (ADAS) applications are aiming at improving both driving safety and comfort. Understanding human drivers' driving styles that make the systems more human-like or personalized for ADAS is the key to improve the system performance, in particular, the acceptance and adaption of ADAS to human drivers. The research presented in this paper focuses on the classification and identification for personalized driving styles. To motivate and reflect the information of different driving styles at the most extent, two sets, which consist of six kinds of stimuli with stochastic disturbance for the leading vehicles are created on a real-time Driver-In-the-Loop Intelligent Simulation Platform (DILISP) with PanoSim-RT®, dSPACE® and DEWETRON® and field test with both RT3000 family and RT-Range respectively.