arXiv subjects

Jan-Pieter Paardekooper

Publications and source records attributed to Jan-Pieter Paardekooper.

At least 19 recordsLinked to original sources

arXiv2026

Towards Safer Autonomous Driving in an Open World: A Dual-Process Approach

Before autonomous driving systems can be deployed on public roads, it is vital that these systems comply with safety standards, traffic rules, and social norms. Although neural networks trained on large amounts of driving data perform well in routine driving tasks, these models often struggle in novel situations that are not well-represented in the data. In this work, we propose a novel framework that combines a neural network for intuitive, learning-based planning in routine driving tasks with model predictive control for reasoning-based planning in unfamiliar situations, inspired by Dual Process Theory. A meta-cognitive component is designed to switch between the two, using a knowledge graph to reason about contextual risk based on explicit perceptual information and relevant traffic rules and social norms. Contextual risk is represented through risk fields, guiding both the switching mechanism in the meta-cognitive component and compliance with safety standards, traffic rules, and social norms in the reasoning-based planner. The effectiveness of our framework is tested in CARLA for variations of a typical out-of-distribution situations involving (emergency) vehicles running a red light. We show that the novel architecture reduces the number of collisions in the scenarios by 89% and improves compliance with the special right-of-way rules, compared to the NN-only planner.

cs.AI↗

arXiv2025

What Did I Learn? Operational Competence Assessment for AI-Based Trajectory Planners

Automated driving functions increasingly rely on machine learning for tasks like perception and trajectory planning, requiring large, relevant datasets. The performance of these algorithms depends on how closely the training data matches the task. To ensure reliable functioning, it is crucial to know what is included in the dataset to assess the trained model's operational risk. We aim to enhance the safe use of machine learning in automated driving by developing a method to recognize situations that an automated vehicle has not been sufficiently trained on. This method also improves explainability by describing the dataset at a human-understandable level. We propose modeling driving data as knowledge graphs, representing driving scenes with entities and their relationships. These graphs are queried for specific sub-scene configurations to check their occurrence in the dataset. We estimate a vehicle's competence in a driving scene by considering the coverage and complexity of sub-scene configurations in the training set. Higher complexity scenes require greater coverage for high competence. We apply this method to the NuPlan dataset, modeling it with knowledge graphs and analyzing the coverage of specific driving scenes. This approach helps monitor the competence of machine learning models trained on the dataset, which is essential for trustworthy AI to be deployed in automated driving.

cs.RO↗

arXiv2023

PRISMA: A Novel Approach for Deriving Probabilistic Surrogate Safety Measures for Risk Evaluation

Surrogate Safety Measures (SSMs) are used to express road safety in terms of the safety risk in traffic conflicts. Typically, SSMs rely on assumptions regarding the future evolution of traffic participant trajectories to generate a measure of risk, restricting their applicability to scenarios where these assumptions are valid. In response to this limitation, we present the novel Probabilistic RISk Measure derivAtion (PRISMA) method. The objective of the PRISMA method is to derive SSMs that can be used to calculate in real time the probability of a specific event (e.g., a crash). The PRISMA method adopts a data-driven approach to predict the possible future traffic participant trajectories, thereby reducing the reliance on specific assumptions regarding these trajectories. Since the PRISMA is not bound to specific assumptions, the PRISMA method offers the ability to derive multiple SSMs for various scenarios. The occurrence probability of the specified event is based on simulations and combined with a regression model, this enables our derived SSMs to make real-time risk estimations. To illustrate the PRISMA method, an SSM is derived for risk evaluation during longitudinal traffic interactions. Since there is no known method to objectively estimate risk from first principles, i.e., there is no known risk ground truth, it is very difficult, if not impossible, to objectively compare the relative merits of two SSMs. Instead, we provide a method for benchmarking our derived SSM with respect to expected risk trends. The application of the benchmarking illustrates that the SSM matches the expected risk trends. Whereas the derived SSM shows the potential of the PRISMA method, future work involves applying the approach for other types of traffic conflicts, such as lateral traffic conflicts or interactions with vulnerable road users.

cs.RO↗

arXiv2023

Connecting the Dots: Context-Driven Motion Planning Using Symbolic Reasoning

The introduction of highly automated vehicles on the public road may improve safety and comfort, although its success will depend on social acceptance. This requires trajectory planning methods that provide safe, proactive, and comfortable trajectories that are risk-averse, take into account predictions of other road users, and comply with traffic rules, social norms, and contextual information. To consider these criteria, in this article, we propose a non-linear model-predictive trajectory generator. The problem space is populated with risk fields. These fields are constructed using a novel application of a knowledge graph, which uses a traffic-oriented ontology to reason about the risk of objects and infrastructural elements, depending on their position, relative velocity, and classification, as well as depending on the implicit context, driven by, e.g., social norms or traffic rules. Through this novel combination, an adaptive trajectory generator is formulated which is validated in simulation through 4 use cases and 309 variations and is shown to comply with the relevant social norms, while taking minimal risk and progressing towards a goal area.

math.OC↗

arXiv2022

Scenario Parameter Generation Method and Scenario Representativeness Metric for Scenario-Based Assessment of Automated Vehicles

The development of assessment methods for the performance of Automated Vehicles (AVs) is essential to enable the deployment of automated driving technologies, due to the complex operational domain of AVs. One candidate is scenario-based assessment, in which test cases are derived from real-world road traffic scenarios obtained from driving data. Because of the high variety of the possible scenarios, using only observed scenarios for the assessment is not sufficient. Therefore, methods for generating additional scenarios are necessary. Our contribution is twofold. First, we propose a method to determine the parameters that describe the scenarios to a sufficient degree without relying on strong assumptions on the parameters that characterize the scenarios. By estimating the probability density function (pdf) of these parameters, realistic parameter values can be generated. Second, we present the Scenario Representativeness (SR) metric based on the Wasserstein distance, which quantifies to what extent the scenarios with the generated parameter values are representative of real-world scenarios while covering the actual variety found in the real-world scenarios. A comparison of our proposed method with methods relying on assumptions of the scenario parametrization and pdf estimation shows that the proposed method can automatically determine the optimal scenario parametrization and pdf estimation. Furthermore, we demonstrate that our SR metric can be used to choose the (number of) parameters that best describe a scenario. The presented method is promising, because the parameterization and pdf estimation can directly be applied to already available importance sampling strategies for accelerating the evaluation of AVs.

cs.RO↗

arXiv2021

Real-World Scenario Mining for the Assessment of Automated Vehicles

Scenario-based methods for the assessment of Automated Vehicles (AVs) are widely supported by many players in the automotive field. Scenarios captured from real-world data can be used to define the scenarios for the assessment and to estimate their relevance. Therefore, different techniques are proposed for capturing scenarios from real-world data. In this paper, we propose a new method to capture scenarios from real-world data using a two-step approach. The first step consists in automatically labeling the data with tags. Second, we mine the scenarios, represented by a combination of tags, based on the labeled tags. One of the benefits of our approach is that the tags can be used to identify characteristics of a scenario that are shared among different type of scenarios. In this way, these characteristics need to be identified only once. Furthermore, the method is not specific for one type of scenario and, therefore, it can be applied to a large variety of scenarios. We provide two examples to illustrate the method. This paper is concluded with some promising future possibilities for our approach, such as automatic generation of scenarios for the assessment of automated vehicles.

cs.RO↗

arXiv2021

Constrained Sampling from a Kernel Density Estimator to Generate Scenarios for the Assessment of Automated Vehicles

The safety assessment of automated vehicles (AVs) is an important aspect of the development cycle of AVs. A scenario-based assessment approach is accepted by many players in the field as part of the complete safety assessment. A scenario is a representation of a situation on the road to which the AV needs to respond appropriately. One way to generate the required scenario-based test descriptions is to parameterize the scenarios and to draw these parameters from a probability density function (pdf). Because the shape of the pdf is unknown beforehand, assuming a functional form of the pdf and fitting the parameters to the data may lead to inaccurate fits. As an alternative, Kernel Density Estimation (KDE) is a promising candidate for estimating the underlying pdf, because it is flexible with the underlying distribution of the parameters. Drawing random samples from a pdf estimated with KDE is possible without the need of evaluating the actual pdf, which makes it suitable for drawing random samples for, e.g., Monte Carlo methods. Sampling from a KDE while the samples satisfy a linear equality constraint, however, has not been described in the literature, as far as the authors know. In this paper, we propose a method to sample from a pdf estimated using KDE, such that the samples satisfy a linear equality constraint. We also present an algorithm of our method in pseudo-code. The method can be used to generating scenarios that have, e.g., a predetermined starting speed or to generate different types of scenarios. This paper also shows that the method for sampling scenarios can be used in case a Singular Value Decomposition (SVD) is used to reduce the dimension of the parameter vectors.

cs.AI↗

arXiv2019

Conditions for Reionizing the Universe with A Low Galaxy Ionizing Photon Escape Fraction

We explore scenarios for reionizing the intergalactic medium with low galaxy ionizing photon escape fractions. We combine simulation-based halo-mass dependent escape fractions with an extrapolation of the observed galaxy rest-ultraviolet luminosity functions to solve for the reionization history from z=20 to z=4. We explore the posterior distributions for key unknown quantities, including the limiting halo mass for star-formation, the ionizing photon production efficiency, and a potential contribution from active galactic nuclei (AGN). We marginalize over the allowable parameter space using a Markov Chain Monte Carlo method, finding a solution which satisfies the most model-independent constraints on reionization. Our fiducial model can match observational constraints with an average escape fraction of <5% throughout the bulk of the epoch of reionization if: i) galaxies form stars down to the atomic cooling limit before reionization and a photosuppression mass of log(M_h/Msol)~9 during/after reionization (-13 -15) dominate the ionizing emissivity, leading to an earlier start to reionization and a smoother evolution of the ionized volume filling fraction than models which assume a single escape fraction at all redshifts and luminosities. The ionizing emissivity from this model is consistent with observations at z=4-5 (and below, when extrapolated), in contrast to some models which assume a single escape fraction. Our predicted ionized volume filling fraction at z=7 of Q_HII=78% (+\- 8%) is in ~1-2 sigma tension with observations of Lya emitters at z~7 and the damping wing analyses of the two known z>7 quasars, which prefer Q_HII,z=7~40-50%.

astro-ph.CO↗

arXiv2017

SPRAI: Coupling of radiative feedback and primordial chemistry in moving mesh hydrodynamics

In this paper we introduce a new radiative transfer code SPRAI (Simplex Photon Radiation in the Arepo Implementation) based on the SimpleX radiation transfer method. This method, originally used only for post-processing, is now directly integrated into the Arepo code and takes advantage of its adaptive unstructured mesh. Radiated photons are transferred from the sources through the series of Voronoi gas cells within a specific solid angle. From the photon attenuation we derive corresponding photon fluxes and ionization rates and feed them to a primordial chemistry module. This gives us a self-consistent method for studying dynamical and chemical processes caused by ionizing sources in primordial gas. Since the computational cost of the SimpleX method does not scale directly with the number of sources, it is convenient for studying systems such as primordial star-forming halos that may form multiple ionizing sources.

astro-ph.IM↗

arXiv2017

Galactic wind X-ray heating of the intergalactic medium during the Epoch of Reionization

The diffuse soft X-ray emissivity from galactic winds is computed during the Epoch of Reionization (EoR). We consider two analytic models, a pressure-driven wind and a superbubble model, and a 3D cosmological simulation including gas dynamics from the First Billion Years (FiBY) project. The analytic models are normalized to match the diffuse X-ray emissivity of star-forming galaxies in the nearby Universe. The cosmological simulation uses physically motivated star formation and wind prescriptions, and includes radiative transfer corrections. The models and the simulation all are found to produce sufficient heating of the Intergalactic Medium to be detectable by current and planned radio facilities through 21 cm measurements during the EoR. While the analytic models predict a 21 cm emission signal relative to the Cosmic Microwave Background sets in by $z_{\rm trans} \simeq 8 - 10$, the predicted signal in the FiBY simulation remains in absorption until reionization completes. The 21 cm absorption differential brightness temperature reaches a minimum of $ΔT \simeq -130$ to $-200$ mK, depending on model. Allowing for additional heat from high mass X-ray binaries pushes the transition to emission to $z_{\rm trans} \simeq 10 - 12$, with shallower absorption signatures having a minimum of $ΔT \simeq -110$ to $-140$ mK. The 21 cm signal may be a means of distinguishing between the wind models, with the superbubble model favouring earlier reheating. While an early transition to emission may indicate X-ray binaries dominate the reheating, a transition to emission as early as $z_{\rm trans} > 12$ would suggest the presence of additional heat sources.

astro-ph.CO↗

arXiv2017

The First Billion Years project: birthplaces of direct collapse black holes

We investigate the environment in which direct-collapse black holes may form by analysing a cosmological, hydrodynamical simulation that is part of the First Billion Years project. This simulation includes the most relevant physical processes leading to direct collapse of haloes, most importantly, molecular hydrogen depletion by dissociation of $H_2$ and $H^-$ from the evolving Lyman-Werner radiation field. We selected a sample of pristine atomic cooling haloes that have never formed stars in their past, have not been polluted with heavy elements and are cooling predominantly via atomic hydrogen lines. Amongst them we identified six haloes that could potentially harbour massive seed black holes formed via direct collapse (with masses in the range of $10^{4-6} M_{sun}$). These potential hosts of direct-collapse black holes form as satellites and are found within 15 physical kpc of proto-galaxies, with stellar masses in the range $10^{5-7} M_{sun}$ and maximal star formation rates of 0.1 Msun/yr over the past 5 Myr, and are exposed to the highest flux of Lyman-Werner radiation emitted from the neighbouring galaxies. It is the proximity to these proto-galaxies that differentiates these haloes from rest of the sample.

astro-ph.CO↗

arXiv2015

The impact of reionization on the formation of supermassive black hole seeds

Direct collapse black holes (DCBHs) formed from the collapse of atomically-cooled primordial gas in the early Universe are strong candidates for the seeds of supermassive BHs. DCBHs are thought to form in atomic cooling haloes in the presence of a strong molecule-dissociating, Lyman-Werner (LW) radiation field. Given that star forming galaxies are likely to be the source of the LW radiation in this scenario, ionizing radiation from these galaxies may accompany the LW radiation. We present cosmological simulations resolving the collapse of primordial gas into an atomic cooling halo, including the effects of both LW and ionizing radiation. We find that in cases where the gas is not self-shielded from the ionizing radiation, the collapse can be delayed by ~ 25 Myr. When the ionized gas does collapse, the free electrons that are present catalyze H2 formation. In turn, H2 cooling becomes efficient in the center of the halo, and DCBH formation is prevented. We emphasize, however, that in many cases the gas collapsing into atomic cooling haloes at high redshift is self-shielding to ionizing radiation. Therefore, it is only in a fraction of such haloes in which DCBH formation is prevented due to reionization.

astro-ph.GA↗

arXiv2015

3D hydrodynamical and radiative transfer modeling of Eta Carinae's colliding winds

We present results of full 3D hydrodynamical and radiative transfer simulations of the colliding stellar winds in the massive binary system Eta Carinae. We accomplish this by applying the SimpleX algorithm for 3D radiative transfer on an unstructured Voronoi-Delaunay grid to recent 3D smoothed particle hydrodynamics (SPH) simulations of the binary colliding winds. We use SimpleX to obtain detailed ionization fractions of hydrogen and helium, in 3D, at the resolution of the original SPH simulations. We investigate several computational domain sizes and Luminous Blue Variable primary star mass-loss rates. We furthermore present new methods of visualizing and interacting with output from complex 3D numerical simulations, including 3D interactive graphics and 3D printing. While we initially focus on Eta Car, the methods employed can be applied to numerous other colliding wind (WR 140, WR 137, WR 19) and dusty 'pinwheel' (WR 104, WR 98a) binary systems. Coupled with 3D hydrodynamical simulations, SimpleX simulations have the potential to help determine the regions where various observed time-variable emission and absorption lines form in these unique objects.

astro-ph.SR↗

arXiv2015

The First Billion Years Project: The escape fraction of ionizing photons in the epoch of reionization

Proto-galaxies forming in low-mass dark matter haloes are thought to provide the majority of ionizing photons needed to reionize the Universe, due to their high escape fractions of ionizing photons. We study how the escape fraction in high-redshift galaxies relates to the physical properties of the halo in which the galaxies form, by computing escape fractions in more than 75000 haloes between redshifts 27 and 6 that were extracted from the First Billion Years project, high-resolution cosmological hydrodynamics simulations of galaxy formation. We find that the main constraint on the escape fraction is the gas column density in a radius of 10 pc around the stellar populations, causing a strong mass dependence of the escape fraction. The lower potential well in haloes with virial mass below 1e8 solar mass results in low column densities that can be penetrated by radiation from young stars (age < 5 Myr). In haloes with virial mass above 1e8 solar mass supernova feedback is important, but only 30% of the haloes in this mass range have an escape fraction higher than 1%. We find a large range of escape fractions in haloes with similar properties, caused by different distributions of the dense gas in the halo. This makes it very hard to predict the escape fraction on the basis of halo properties and results in a highly anisotropic escape fraction. The strong mass dependence, the large spread and the large anisotropy of the escape fraction may strongly affect the topology of reionization and is something current models of cosmic reionization should strive to take into account.

astro-ph.CO↗

arXiv2015

3D radiative transfer simulations of Eta Carinae's inner colliding winds - II: Ionization structure of helium at periastron

Spectral observations of the massive colliding wind binary Eta Carinae show phase-dependent variations, in intensity and velocity, of numerous helium emission and absorption lines throughout the entire 5.54-year orbit. Approaching periastron, the 3D structure of the wind-wind interaction region (WWIR) gets highly distorted due to the eccentric ($e \sim 0.9$) binary orbit. The secondary star ($η_{\mathrm{B}}$) at these phases is located deep within the primary's dense wind photosphere. The combination of these effects is thought to be the cause of the particularly interesting features observed in the helium lines at periastron. We perform 3D radiative transfer simulations of $η$ Car's interacting winds at periastron. Using the SimpleX radiative transfer algorithm, we post-process output from 3D smoothed particle hydrodynamic simulations of the inner 150 au of the $η$ Car system for two different primary star mass-loss rates ($\dot{M}_{η_{\mathrm{A}}}$). Using previous results from simulations at apastron as a guide for the initial conditions, we compute 3D helium ionization maps. We find that, for higher $\dot{M}_{η_{\mathrm{A}}}$, $η_{\mathrm{B}}$ He$^{0+}$-ionizing photons are not able to penetrate into the pre-shock primary wind. He$^{+}$ due to $η_{\mathrm{B}}$ is only present in a thin layer along the leading arm of the WWIR and in a small region close to the stars. Lowering $\dot{M}_{η_{\mathrm{A}}}$ allows $η_{\mathrm{B}}$'s ionizing photons to reach the expanding unshocked secondary wind on the apastron side of the system, and create a low fraction of He$^{+}$ in the pre-shock primary wind. With apastron on our side of the system, our results are qualitatively consistent with the observed variations in strength and radial velocity of $η$ Car's helium emission and absorption lines, which helps better constrain the regions where these lines arise.

astro-ph.SR↗

arXiv2015

3D Printing Meets Computational Astrophysics: Deciphering the Structure of Eta Carinae's Inner Colliding Winds

We present the first 3D prints of output from a supercomputer simulation of a complex astrophysical system, the colliding stellar winds in the massive (>120 M_Sun), highly eccentric (e ~ 0.9) binary star system Eta Carinae. We demonstrate the methodology used to incorporate 3D interactive figures into a PDF journal publication and the benefits of using 3D visualization and 3D printing as tools to analyze data from multidimensional numerical simulations. Using a consumer-grade 3D printer (MakerBot Replicator 2X), we successfully printed 3D smoothed particle hydrodynamics (SPH) simulations of Eta Carinae's inner (r ~ 110 au) wind-wind collision interface at multiple orbital phases. The 3D prints and visualizations reveal important, previously unknown 'finger-like' structures at orbital phases shortly after periastron (phi ~ 1.045) that protrude radially outward from the spiral wind-wind collision region. We speculate that these fingers are related to instabilities (e.g. thin-shell, Rayleigh-Taylor) that arise at the interface between the radiatively-cooled layer of dense post-shock primary-star wind and the fast (3000 km/s), adiabatic post-shock companion-star wind. The success of our work and easy identification of previously unrecognized physical features highlight the important role 3D printing and interactive graphics can play in the visualization and understanding of complex 3D time-dependent numerical simulations of astrophysical phenomena.

astro-ph.SR↗

arXiv2014

3D radiative transfer simulations of Eta Carinae's inner colliding winds - I. Ionization structure of helium at apastron

The highly eccentric binary system Eta Carinae shows numerous time-variable emission and absorption features. These observational signatures are the result of interactions between the complex three-dimensional (3D) wind-wind collision regions and photoionization by the luminous stars. Specifically, helium presents several interesting spectral features that provide important clues on the geometry and physical properties of the system and the individual stars. We use the SimpleX algorithm to post-process 3D smoothed particle hydrodynamics simulation output of the interacting winds in Eta Car in order to obtain the fractions of ionized helium assuming three different primary star mass-loss rates. The resultant ionization maps constrain the regions where helium is singly- and doubly-ionized. We find that reducing the primary's mass-loss rate increases the volume of He+. Lowering the primary mass-loss rate produces large variations in the volume of He+ in the pre-shock primary wind on the periastron side of the system. Our results show that binary orientations in which apastron is on our side of the system are more consistent with available observations. We suggest that small variations in the primary's mass-loss rate might explain the observed increase in HeI absorption in recent decades, although numerous questions regarding this scenario remain open. We also propose that the absence of broad HeI lines in the spectra of Eta Car between its 1890's eruption and 1944 might be explained by the companion star's He0+ ionizing photons not being able to penetrate the wind-wind interaction region, due to a higher primary mass-loss rate at that time (by a factor >2, compared to the present value).

astro-ph.SR↗

arXiv2012

The First Billion Years project - IV: Proto-galaxies reionising the Universe

The contribution of stars in galaxies to cosmic reionisation depends on the star formation history in the Universe, the abundance of galaxies during reionisation, the escape fraction of ionising photons and the clumping factor of the inter-galactic medium (IGM). We compute the star formation rate and clumping factor during reionisation in a cosmological volume using a high-resolution hydrodynamical simulation. We post-process the output with detailed radiative transfer simulations to compute the escape fraction of ionising photons. Together, this gives us the opportunity to assess the contribution of galaxies to reionisation self-consistently. The strong mass and redshift dependence of the escape fraction indicates that reionisation occurred between z=15 and z=10 and was mainly driven by proto-galaxies forming in dark-matter haloes with masses between 1e7 and 1e8 solar mass. More massive galaxies that are rare at these redshifts and have significantly lower escape fractions contribute less photons to the reionisation process than the more-abundant low-mass galaxies. Star formation in the low-mass haloes is suppressed by radiative feedback from reionisation, therefore these proto-galaxies only contribute when the part of the Universe they live in is still neutral. After z~10, massive galaxies become more abundant and provide most of the ionising photons. In addition, we find that Population (Pop) III stars are too short-lived and not frequent enough to have a major contribution to reionisation. Although the stellar component of the proto-galaxies that produce the bulk of ionising photons during reionisation is too faint to be detected by the James Webb Space Telescope (JWST), these sources are brightest in the H-alpha and Ly-alpha recombination lines, which will likely be detected by JWST in deep surveys.

astro-ph.CO↗