Lars Hammarstrand

Road Intensity Based Mapping Using Radar Measurements With a Probability Hypothesis Density Filter

Mapping stationary objects is essential for autonomous vehicles and many autonomous functions in vehicles. In this contribution the probability hypothesis density (PHD) filter framework is applied to automotive imagery sensor data for constructing such a map, where the main advantages are that it avoids the detection, the data association and the track handling problems in conventional multiple-target tracking, and that it gives a parsimonious representation of the map in contrast to grid based methods. Two original contributions address the inherent complexity issues of the algorithm: First, a data clustering algorithm is suggested to group the components of the PHD into different clusters, which structures the description of the prior and considerably improves the measurement update in the PHD filter. Second, a merging step is proposed to simplify the map representation in the PHD filter. The algorithm is applied to multi-sensor radar data collected on public roads, and the resulting map is shown to well describe the environment as a human perceives it.

A Probabilistic Framework for Decision-Making in Collision Avoidance Systems

This paper is concerned with the problem of decision-making in systems that assist drivers in avoiding collisions. An important aspect of these systems is not only assisting the driver when needed but also not disturbing the driver with unnecessary interventions. Aimed at improving both of these properties, a probabilistic framework is presented for jointly evaluating the driver acceptance of an intervention and the necessity thereof to automatically avoid a collision. The intervention acceptance is modeled as high if it estimated that the driver judges the situation as critical, based on the drivers observations and predictions of the traffic situation. One advantage with the proposed framework is that interventions can be initiated at an earlier stage when the estimated driver acceptance is high. Using a simplified driver model, the framework is applied to a few different types of collision scenarios. The results show that the framework has appealing properties, both with respect to increasing the system benefit and to decreasing the risk of unnecessary interventions.

Extended Object Tracking using a Radar Resolution Model

The work presented here concerns the problem of vehicle tracking when multiple radar reflection centers could be resolved on each vehicle. For this extended target tracking problem we propose a radar sensor model, capable of describing such measurements, incorporating sensor resolution. Furthermore, we introduce approximations to handle the inherently complex data association problem. The evaluation in terms of describing measured data and resulting tracking performance shows that the model effectively exploits the information in multiple vehicle detections.

A New Vehicle Motion Model for Improved Predictions and Situation Assessment

Reliable and accurate vehicle motion models are of vital importance for automotive active safety systems for a number of reasons. First of all, these models are necessary in tracking algorithms that provide the safety system with information. Second, the motion model is often used by the safety application to make long-term predictions about the future traffic situation. These predictions are then part of the basic data used by the system to determine if, when, and how to intervene. In this paper, we suggest a framework for designing accurate vehicle motion models. The resulting models differ from conventional models in that the expected control input from the driver is included. By also providing a methodology for a formal treatment of the uncertainties, a model structure well suited, e.g., in a tracking algorithm, is obtained. To utilize the framework in an application will require careful design and validation of submodels to calculate the expected driver control input. We illustrate the potential of the framework by examining the performance for a specific model example using real measurements. The properties are compared with those of a constant acceleration model. Evaluations indicate that the proposed model yields better predictions and that it has an ability to estimate the prediction uncertainties.

Bayesian Road Estimation Using Onboard Sensors

This paper describes an algorithm for estimating the road ahead of a host vehicle based on the measurements from several onboard sensors: a camera, a radar, wheel speed sensors, and an inertial measurement unit. We propose a novel road model that is able to describe the road ahead with higher accuracy than the usual polynomial model. We also develop a Bayesian fusion system that uses the following information from the surroundings: lane marking measurements obtained by the camera and leading vehicle and stationary object measurements obtained by a radar-camera fusion system. The performance of our fusion algorithm is evaluated in several drive tests. As expected, the more information we use, the better the performance is.

Vehicle self-localization using off-the-shelf sensors and a detailed map

In the research on autonomous vehicles, self-localization is an important problem to solve. In this paper we present a localization algorithm based on a map and a set of off-the-shelf sensors, with the purpose of evaluating this low-cost solution with respect to localization performance. The used test vehicle is equipped with a Global Positioning System receiver, a gyroscope, wheel speed sensors, a camera providing information about lane markings, and a radar detecting landmarks along the road. Evaluation shows that the localization result is within or close to the requirements for autonomous driving when lane markers and good radar landmarks are present. However, it also indicates that the solution is not robust enough to handle situations when one of these information sources is absent.

Multitarget Sensor Resolution Model and Joint Probabilistic Data Association

In the design of target tracking algorithms, the aspect of sensor resolution is rarely considered. Instead, it is usually assumed that all targets are always resolved, and that the only uncertainties in the data association are which targets that are detected, and which measurement each detected target gave rise to. However, in situations where the targets are closely spaced in relation to the sensor resolution, this assumption is not valid, and may lead to degraded tracking performance due to an incorrect description of the data. We present a framework for handling sensor resolution effects for an arbitrary, but known, number of targets. We propose a complete multitarget sensor resolution model that can be incorporated into traditional Bayesian tracking filters. Further, the exact form of the posterior probability density function (pdf) is derived, and two alternative ways of approximating that exact posterior density with a joint probabilistic data association (JPDA) filter are proposed. Evaluations of the resulting filters on simulated radar data show significantly increased tracking performance compared with the JPDA filter without a resolution model.

Adaptive Radar Sensor Model for Tracking Structured Extended Objects

We propose a tracking framework jointly estimating the position of a single extended object and the set of radar reflectors that it contains. The reflectors are assumed to lie on a line structure, but the number of reflectors and their positions on the line are unknown. Additionally, we incorporate an accurate radar sensor model considering the resolution capabilities of the sensor. The evaluation of the framework on radar measurements shows promising results.

A CPHD Filter for Tracking With Spawning Models

In some applications of multi-target tracking, appearing targets are suitably modeled as spawning from existing targets. However, in the original formulation of the cardinalized probability hypothesis density (CPHD) filter, this type of model is not supported; instead appearing targets are modeled by spontaneous birth only. In this paper we derive the necessary equations for a CPHD filter for the case when the process model also includes target spawning. For this generalized filter, the cardinality prediction formula might become computationally intractable for general spawning models. However, when the cardinality distribution of the spawning targets is either Bernoulli or Poisson, we derive expressions that are practical and computationally efficient. Simulations show that the proposed filter responds faster to a change in target number due to spawned targets than the original CPHD filter. In addition, the performance of the filter, considering the optimal subpattern assignment (OSPA), is improved when having an explicit spawning model.

Road geometry estimation using a precise clothoid road model and observations of moving vehicles

An important part of any advanced driver assistance system is road geometry estimation. In this paper, we develop a Bayesian estimation algorithm using lane marking measurements received from a camera and measurements of the leading vehicles received from a radar-camera fusion system, to estimate the road up to 200 meters ahead in highway scenarios. The filtering algorithm uses a segmented clothoid-based road model. In order to use the heading of leading vehicles we need to detect if each vehicle is keeping lane or changing lane. Hence, we propose to jointly detect the motion state of the leading vehicles and estimate the road geometry using a multiple model filter. Finally the proposed algorithm is compared to an existing method using real data collected from highways. The results indicate that it provides a more accurate road estimation in some scenarios.