arXiv subjects

Gabriel Hartmann

Publications and source records attributed to Gabriel Hartmann.

5 recordsLinked to original sources

arXiv2026

Testing MOND-like modifications to gravity using growth-rate measurements and one-loop corrections to the matter power spectrum

We develop a perturbative framework for structure formation in a broad class of MOND-like theories characterized by a generalized nonlinear Poisson equation. We derive the modified evolution equations governing matter perturbations and obtain the corresponding linear growth equation, extending the analysis into the mildly nonlinear regime through one-loop corrections to the matter power spectrum. Beyond the theoretical framework, we perform a cosmological analysis based on two phenomenological scenarios: one parametrized by a quantity controlling the degree of nonlinearity in the generalized Poisson equation, and another describing the interplay between the MOND acceleration scale and the cosmological acceleration associated with the background expansion. We constrain these scenarios using recent measurements of the growth rate of structure, $f\sigma_8$, DESI-DR2 baryon acoustic oscillation data, and Type Ia supernova compilations. We find no statistically significant evidence for departures from the standard $\Lambda$CDM cosmology. The inferred constraints are fully consistent with the GR + $\Lambda$CDM scenario within the current observational uncertainties. At nonlinear scales, we note that MOND-like modifications can alter $P_{\rm NL}(k)$ and leave signatures that current and future high-precision large-scale-structure observations may probe. Our results establish a systematic connection between MOND-like gravitational dynamics and large-scale structure observations, providing a consistent framework to assess the phenomenological viability of MOND-inspired modifications of gravity in a cosmological context.

astro-ph.CO↗

arXiv2025

One-loop power spectrum corrections in interacting dark energy cosmologies

Interacting Dark Energy (IDE) models offer a promising avenue to explore possible exchanges of energy and momentum between dark matter and dark energy, providing a dynamical extension of the standard $\Lambda$CDM paradigm. Such interactions modify the growth of cosmic structures, imprinting distinctive signatures on the matter power spectrum that can be tested through large-scale structure (LSS) observations. In this work, we compute the one-loop corrections to the matter power spectrum in IDE models. We then reinterpret these results within the standard framework of the Effective Field Theory of Large-Scale Structure (EFTofLSS), which provides a consistent description of mildly non-linear scales and allows for reliable comparisons with observational data. We investigate two commonly studied forms of the coupling function, $Q$, namely $Q = \xi \mathcal{H} \rho_{\rm m}$ and $Q = \xi \mathcal{H} \rho_{\rm DE}$, and introduce a novel interaction term, $Q = \Gamma \, \rho_{\rm m} \, \rho_{\rm DE} \, \theta_{\rm m}$, characterized by the non-linear coupling constant $\Gamma$, which links the interaction strength to the velocity divergence of dark matter. This coupling function is proposed to isolate the effects solely of the IDE model on mildly non-linear scales. Using Full-Shape (FS) measurements of the galaxy power spectrum from BOSS DR12, we constrain the interaction rate $\Gamma$, the cosmological parameters, and the bias parameters. We find $\Gamma = 0.0039 \pm 0.0082$, which is highly consistent with the $\Lambda$CDM model. This work opens the possibility of testing IDE models at mildly non-linear scales, potentially providing new insights for this class of models beyond the standard $\Lambda$CDM framework.

astro-ph.CO↗

arXiv2022

Meta-Reinforcement Learning Using Model Parameters

In meta-reinforcement learning, an agent is trained in multiple different environments and attempts to learn a meta-policy that can efficiently adapt to a new environment. This paper presents RAMP, a Reinforcement learning Agent using Model Parameters that utilizes the idea that a neural network trained to predict environment dynamics encapsulates the environment information. RAMP is constructed in two phases: in the first phase, a multi-environment parameterized dynamic model is learned. In the second phase, the model parameters of the dynamic model are used as context for the multi-environment policy of the model-free reinforcement learning agent.

cs.LG↗

arXiv2021

Competitive Driving of Autonomous Vehicles

This paper describes Ariel Team's autonomous racing controller for the Indy Autonomous Challenge (IAC) simulation race. IAC is the first multi-vehicle autonomous head-to-head competition, reaching speeds of 300 km/h along an oval track, modeled after the Indianapolis Motor Speedway (IMS). Our racing controller attempts to maximize progress along the track while avoiding collisions with opponent vehicles and obeying the race rules. To this end, the racing controller first computes a race line offline. Then, it repeatedly computes online a small set of dynamically feasible maneuver candidates, each tested for collision with the opponent vehicles. Finally, it selects the maneuver that maximizes progress along the track, taking into account the race line. The maneuver candidates, as well as the predicted trajectories of the opponent vehicles, are approximated using a point mass model. Despite the simplicity of this racing controller, it managed to drive competitively and with no collision with any of the opponent vehicles in the IAC final simulation race.

cs.RO↗

arXiv2018

Deep Reinforcement Learning for Time Optimal Velocity Control using Prior Knowledge

Autonomous navigation has recently gained great interest in the field of reinforcement learning. However, little attention was given to the time optimal velocity control problem, i.e. controlling a vehicle such that it travels at the maximal speed without becoming dynamically unstable (roll-over or sliding). Time optimal velocity control can be solved numerically using existing methods that are based on optimal control and vehicle dynamics. In this paper, we use deep reinforcement learning to generate the time optimal velocity control. Furthermore, we use the numerical solution to further improve the performance of the reinforcement learner. It is shown that the reinforcement learner outperforms the numerically derived solution, and that the hybrid approach (combining learning with the numerical solution) speeds up the training process.

cs.RO↗