Mohamed Essayed Bouzouraa

Fusion of occupancy grid mapping and model based object tracking for driver assistance systems using laser and radar sensors

in this paper we present a novel environment perception system based on an occupancy grid mapping and a multi-object tracking. The goal of such a system is to create a harmonic, consistent and complete representation of the vehicle environment as a base for future advanced driver assistance systems. In addition to a mathematical formulation of the problem we present a robust algorithm to detect dynamic obstacles from the occupancy map and show how both, the mapping process and the tracking can benefit from each other. Therefore, the concept of moving objects with associated dynamic cells is introduced. The presented techniques are applicable to both 2D and 3D mapping and can be also extended to correct the ego motion from the occupancy map and the object tracks. Unlike many publications over the last years our work provides real time performance and an accurate detection of obstacles with real laser and radar sensors and can fulfill the requirements of future driver assistance systems.

Road curb detection based on different elevation mapping techniques

A road curb detection algorithm for a 3D sensor, e.g. a dense stereo camera, is presented in this paper. The road curb detection is based on a digital elevation map. Different techniques and coordinate systems for mapping the height values are compared theoretically and by simulating a different quality of ego motion data. Furthermore we introduce a new approach of finding road curbs in an elevation map, which is based on a calculation of the most probable path. Using an elevation map the curb height can be calculated in an additional step. For evaluation we use highly accurate reference sensors and compare the detected curbs to a ground truth. Additionally we introduce a novel criteria to describe the quality of an elevation map and discuss the results. The road detection algorithm works in real-time and has a position accuracy of about 10 cm and an height error of about 1.5 cm.

EgoMaster: A central ego motion estimation for driver assist systems

In this paper we present an approach for a central ego motion estimation with standard and near-series sensors for advanced driver assist systems. The two main contributions of this article are the provision of all variables relative to the road surface and the fusion of several sensors and perception modules providing information of ego localization. Thus we employ a discrete Kalman filter to consider the standard lift sensors of the suspension and to take the radial tire deflection into account. Additionally, a reference sensor set is introduced which enables to evaluate all measured and estimated states, even those which are not related to the vector of gravity. The validation by the reference sensor set demonstrates the accurrancy and the efficiency of the approach.

Multi-cue, Model-Based Detection and Mapping of Road Curb Features Using Stereo Vision

In this paper a new approach for road curb detection is presented, which is able to detect road curbs independently of their geometry and orientation in relation to the car. This approach includes a novel algorithm for curb feature detection, which combines cues for road curbs from a 3D point cloud and an intensity image, both gained from a stereo camera system. The edge detection operators used for the detection of curb features are adapted according to curb models. With these adaptive operators road curbs can be found independently from their position and orientation in the image and the correct curb height can be calculated. Further we introduce a computation efficient grid map, which is based on a local linear description of the road curb in every cell and is able to represent the road curbs three-dimensionally. A chaining mechanism, which benefits from the estimated curb directions in the grid map, generates road curbs represented as 3D polygonal chains. An evaluation of the detected curbs using DGPS measurements of the curbs and the car shows a high position and height accuracy for this real time capable approach.

Circle Detection in Single-layer Laser Scans for Roundabout Perception

This work presents a circle detection algorithm for the purpose of detecting roundabout central islands in single-layer laser scans. The algorithm is robust against measurement errors and imperfections of the central islands circular form and it can be used for laser scanners with only a few vertical layers. Main features of the algorithm are the semi-convex segmentation and curvature-based segmentation steps and the use of an a priori digital map to classify the detected circles. Experimental results show that, based on its performance, the method can be integrated in a real-time online roundabout perception system for highly automated vehicles.

Base Algorithms of Environment Maps and Efficient Occupancy Grid Mapping on Embedded GPUs.

An accurate model of the environment is essential for future Advanced Driver Assistance Systems (ADASs). To generate such a model, an enormous amount of data has to be fused and processed. Todays Electronic Control Units (ECUs) struggle to provide enough computing power for those future tasks. To overcome these shortcomings, new architectures, like embedded Graphics Processing Units (GPUs), have to be introduced. For future ADASs, also sensors with a higher accuracy have to be used. In this paper, we analyze common base algorithms of environment maps based on the example of the occupancy grid map. We show from which sensor resolution it is rational to use an (embedded) GPU and which speedup can be achieved compared to a Central Processing Unit (CPU) implementation. A second contribution is a novel method to parallelize an occupancy grid map on a GPU, which is computed from the sensor values of a lidar scanner with several layers. We evaluate our introduced algorithm with real driving data collected on the autobahn.

Convoy tracking for ADAS on embedded GPUs

Future Advanced Driver Assistance Systems (ADAS) need to create an accurate model of the environment. Accordingly, an enormous amount of data has to be fused and processed. From this data, information such as the positions of the vehicles, has to be extracted out of the model, e.g., to create a convoy track. Common architectures used today, like single-core processors in automotive Electronic Control Units (ECUs), struggle to provide enough computing power for those tasks. Here, emerging embedded multi-core architectures are appealing such as embedded Graphics Processing Units (GPUs). In this paper, we present a novel parallelization of a convoy track detection algorithm. Moreover, in order to profit best from for embedded GPUs, special techniques such as Zero Copy are exploited to parallelize our application. As an experimental platform, an Nvidia Tegra K1 is used, which is also common in the automotive industry. For different scenarios, we illustrate the limitations of the system and algorithm. Yet, impressive speedups with respect to a single-core CPU solution of up to nine may be achieved using the proposed parallelization techniques in case of high traffic situations.

Towards a Multi-hypothesis Road Representation for Automated Driving

The process of perception inevitably involves uncertainty. In the field of automated driving, however, uncertainty about the roadway - and especially about the understanding of the roadway - is yet mostly neglected. In this paper it is argued that this uncertainty should be represented and considered in the process of behavior generation. A road representation capable of representing multiple hypotheses about the roadway is presented. The representation language allows to express where and how hypotheses differ and to infer this information in an efficient way. The focus of this paper is on the qualitative representation of uncertainty.

Laser scanner based road surface estimation for automotive applications

Driver Assistance Systems and Active Suspension Systems require the accurate perception of the vehicles environment. In the case of Active Suspension Systems the road surface in front of the wheels has to be reconstructed with an accuracy of few centimeters. In this work we present a new method to generate the relevant road height information based on an automotive compact laser scanner mounted in the cars radiator grill and originally designed for Driver Assistance Systems. Our method is based on the generation of a statistical and a deterministic sensor model which permits to precisely extract the ground height from the raw distance measurements. Additionally, the concept of the corridor representation is introduced. It efficiently processes the 3D data in the relevant areas and integrates past measurements enabling for the first time the representation of long road waves despite limited sensor range. Experimental results based on the comparison of the output of our method with the ground truth of a roadway ramp show the suitability of the presented work.