Michael Maile

Balancing the requirements for a zero false positive/negative Forward Collision Warning

The main benefit of cooperative vehicular communication networks is the increased traffic safety that safety applications could introduce. Although the technological aspects have been extensively researched, the feasibility regions in which communication could support safety applications are yet not clear. This is, primarily, due to the lack of clearly defined application requirements. Moreover, reduction of false positives and false negatives are either not considered at all or only partially. In the current paper we investigate under which conditions vehicular communication can support safety applications. For this, we identified the requirements of safety applications, on the example of Forward Collision Warning, first from application perspective and pursuing zero false positive and zero false negative constraint. Afterwards, we quantify what is the maximum vehicle density for which application requirements can be supported under realistic communication conditions, as well as how much the idealistic application requirements have to be relaxed in order to achieve a balance between scalability and zero false rates.

Intersection Collision Avoidance: From Driver Alerts to Vehicle Control

Intersection Collision Avoidance (ICA) based on Dedicated Short Range Communications (DSRC) is one of the most promising applications for vehicle communications. System implementations based on vehicle sensors suffer from field of view limitations that vehicle communications do not exhibit. This makes particularly interesting the adoption of DSRC in combination with other onboard sensors to address intersection crash scenarios. State-of-the-art DSRC-based intersection collision avoidance systems, notably the Intersection Movement Assist (IMA) application, are aimed to issue audible and visual alerts to the driver. This paper describes an implementation of Intersection Collision Avoidance that extends the IMA concept to cover a wider range of intersection collision scenarios and introduces automated braking as a system response when a high risk of a collision is detected and the driver does not react to the alerts.

A random finite set approach for dynamic occupancy grid maps with real-time application

Grid mapping is a well-established approach for environment perception in robotic and automotive applications. Early work suggests estimating the occupancy state of each grid cell in a robot’s environment using a Bayesian filter to recursively combine new measurements with the current posterior state estimate of each grid cell. This filter is often referred to as binary Bayes filter. A basic assumption of classical occupancy grid maps is a stationary environment. Recent publications describe bottom-up approaches using particles to represent the dynamic state of a grid cell and outline prediction-update recursions in a heuristic manner. This paper defines the state of multiple grid cells as a random finite set, which allows to model the environment as a stochastic, dynamic system with multiple obstacles, observed by a stochastic measurement system. It motivates an original filter called the probability hypothesis density / multi-instance Bernoulli (PHD/MIB) filter in a top-down manner. The paper presents a real-time application serving as a fusion layer for laser and radar sensor data and describes in detail a highly efficient parallel particle filter implementation. A quantitative evaluation shows that parameters of the stochastic process model affect the filter results as theoretically expected and that appropriate process and observation models provide consistent state estimation results.

Extended object tracking using IMM approach for a real-world vehicle sensor fusion system

Autonomous driving poses unique challenges for vehicle environment perception due to the complicated driving environment where the autonomous vehicle connects itself with surrounding objects. Precise tracking of the relevant dynamic traffic participants (e.g., vehicle/byciclist/pedestrian) becomes a key component for the task of comprehensive environmental perception and reliable scene understanding. It is necessary for vehicle trackers to treat the objects as extended (rigid) target, as opposed to traditional point target tracking (say, in aerospace applications). The extended object tracking is an extremely challenging problem in real world, due to high requirements of the object estimation on accuracy of kinematic/shape information, association robustness, model match on various target motion behaviors, and statistical property amicability (e.g., estimation consistency/covariance reliability). We present an extended object tracker — based on an interacting multiple model with unbiased mixing estimator for kinematic information at a specified tracking reference point, a truncated Gaussian scheme for shape (width/length/orientation) estimation, and a hierarchical association method according to both kinematic and shape information — to tackle all of the major challenges. Our special effort is put on handling an intriguing conflict between theory and practice: the so-called likelihood credibility issue. That is, the likelihood is expected to credibly reflect the data statistical probability but is actually distorted/drifting in real world systems, due to mainly artificial physics introduced in multiple-stage data processing. In this study, from systematic point of view, we design an interacting multiple model based extended object tracker with proper likelihood compensation in the statistically-distorted real world. It can be shown that the presented tracker can deliver an effective estimation performance in real road traffic of the imperfect world.

Object Matching for Inter-Vehicle Communication Systems—An IMM-Based Track Association Approach With Sequential Multiple Hypothesis Test

Autonomous driving poses unique challenges for vehicle environment perception due to the complex driving environment where the autonomous vehicle interacts with surrounding traffic participants. Due to the limited capability of any sensor perception system, it is highly desirable that an autonomous driving vehicle could use not only information from onboard sensors (say, radar/camera/lidar) but also from remote (network) information via inter-vehicle communication systems. The collaborative information from cooperative/non-cooperative remote vehicles (along with the onboard sensor data) could substantially improve the vehicle decision making process and push autonomous driving to be safer and more reliable. Inter-vehicle communication technologies are at the stage of development for market introduction, after years of research and standardization. In this paper, we setup a dedicated short range communication (DSRC) system to provide a low-latency inter-vehicle wireless communication channel. The task is to build a record linkage between the onboard sensor data and the corresponding DSRC-transmitted remote vehicle information when both sets belong to the same object, for the purpose of enhancing host vehicle environment perceiving capability and reliability. This is a typical data association problem. The challenges mainly lie in the inherent uncertain nature of the observation data and the practical issues that information often suffers from delays and drops. We propose a track-based association approach using an interacting multiple model estimator with a sequential multiple hypothesis test (denoted as IMM-SMHT) as an ubiquitous solution to handle different situations in complicated driving scenarios. To fully exploit the potential of such a system, only position information (from the DSRC channel and onboard radar system) is used for the object matching purpose—we try to use the least amount of information to achieve a high association accuracy; additional information can be used but not currently considered. We aim to provide a real world solution, and therefore, a prototype vehicle system is built with practical consideration on market availability, cost, and sensor limitations. We design meaningful use cases for creating functionality modules from a systematic point of view. The inter-vehicle information fusion system based on the track fusion approach using the IMM-SMHT is tested in real traffic on the U.S. roads and shows promising object matching performance of significant practical feasibility.

Object tracking with de-autocorrelation scheme for a dynamic occupancy gridmap system

Autonomous driving poses unique challenges for vehicle environment perception in complex driving environments. Due to the uncertain nature of the vehicle environment and imperfection of any perception framework, multiple stages of estimation might be necessary to achieve the desired performance. However, it is highly possible that the estimation of one stage might result in output estimates with significant auto/cross-correlation, which would pass to another stage. In such situations, a decorrelation procedure is required. We present an object tracking approach taking into consideration the auto-correlation (of the velocity components) introduced by an upfront dynamic occupancy gridmap system. More specifically, we use a linear state-space system to approximately “reconstruct” the nonlinear estimation/mapping procedure for the purpose of auto-correlation quantification. We focus on demonstrating an estimation improvement of the proposed decorrelation tracker over a direct Kalman filter (i.e., KF that ignores the auto-correlation). For this we recorded several scenarios of different target motion behaviors. It is shown that the decorrelation tracker indeed introduces noticable estimation improvement in particularly velocity space if the target moves in a relatively smooth manner.

Fundamental properties of dynamic occupancy grid systems for vehicle environment perception

Using probabilistic reasoning to model and perceive driving environments is a challenging problem due to both geometric and dynamic random natures. The occupancy grid system, providing an intermediate representation for complicated environments and having many beneficial implications (such as avoiding direct data association and more freedom in data fusion), has been increasingly becoming a popular paradigm for vehicle environment perception. The conventional static occupancy grid (SOG) system only describes static environments; to incorporate dynamic information into the conventional occupancy grid is naturally desirable. The corresponding system is called dynamic occupancy grid (DOG). The DOG systems are still developing and some fundamental questions remain unanswered. In a statistical sense, the DOG extends the SOG in generalizing its grid as a random field defined over a parameter space in not only geometric space but also time domain. In this paper, under a formal statistical definition of the DOG, we carry out a Bayesian analysis and examine commonly-used assumptions and approximations in the literature. In view of works having been done, we present a particle-based multiple model approach for our DOG system and the corresponding results are given in a typical vehicle driving scenario.

Study of the microdoppler signature of a bicyclist for different directions of approach

The successful implementation of autonomous driving in an urban setting depends on the ability of the environment perception system to correctly classify vulnerable road users such as pedestrians and bicyclists in dense, complex scenarios. Self-driving vehicles include sensor systems such as cameras, lidars, and radars to enable decision making. Among these systems, radars are particularly relevant due to their operational robustness under adverse weather and night light conditions. Classification of pedestrian and car in urban settings using automotive radar has been widely investigated, suggesting that micro-Doppler signatures are useful for target discrimination. Our objective is to analyze and study the micro-Doppler signature of bicyclists approaching a vehicle from different directions in order to establish the basis of a classification criterion to distinguish bicycles from other targets including clutter. The micro-Doppler signature is obtained by grouping individual reflecting points using a clustering algorithm and observing the evolution of all the points belonging to an object in the Doppler domain over time. A comparison is then made with simulated data that uses a kinematic model of bicyclists’ movement. The suitability of the micro-Doppler bicyclist signature as a classification feature is determined by comparing it to those belonging to cars and pedestrians approaching the automotive radar system.