Searcharxiv⌕ Search

arXiv subjects

George Papadopoulos

Publications and source records attributed to George Papadopoulos.

At least 19 recordsLinked to original sources
arXiv2026

Learning to maintain safety through expert demonstrations in settings with unknown constraints: A Q-learning perspective

Given a set of trajectories demonstrating the execution of a task safely in a constrained MDP with observable rewards but with unknown constraints and non-observable costs, we aim to find a policy that maximizes the likelihood of demonstrated trajectories trading the balance between being conservative and increasing significantly the likelihood of high-rewarding trajectories but with potentially unsafe steps. Having these objectives, we aim towards learning a policy that maximizes the probability of the most $promising$ trajectories with respect to the demonstrations. In so doing, we formulate the ``promise" of individual state-action pairs in terms of $Q$ values, which depend on task-specific rewards as well as on the assessment of states' safety, mixing expectations in terms of rewards and safety. This entails a safe Q-learning perspective of the inverse learning problem under constraints: The devised Safe $Q$ Inverse Constrained Reinforcement Learning (SafeQIL) algorithm is compared to state-of-the art inverse constraint reinforcement learning algorithms to a set of challenging benchmark tasks, showing its merits.

cs.LG↗
arXiv2025

Learning safe, constrained policies via imitation learning: Connection to Probabilistic Inference and a Naive Algorithm

This article introduces an imitation learning method for learning maximum entropy policies that comply with constraints demonstrated by expert trajectories executing a task. The formulation of the method takes advantage of results connecting performance to bounds for the KL-divergence between demonstrated and learned policies, and its objective is rigorously justified through a connection to a probabilistic inference framework for reinforcement learning, incorporating the reinforcement learning objective and the objective to abide by constraints in an entropy maximization setting. The proposed algorithm optimizes the learning objective with dual gradient descent, supporting effective and stable training. Experiments show that the proposed method can learn effective policy models for constraints-abiding behaviour, in settings with multiple constraints of different types, accommodating different modalities of demonstrated behaviour, and with abilities to generalize.

cs.LG↗
arXiv2025

An Extended Benchmarking of Multi-Agent Reinforcement Learning Algorithms in Complex Fully Cooperative Tasks

Multi-Agent Reinforcement Learning (MARL) has recently emerged as a significant area of research. However, MARL evaluation often lacks systematic diversity, hindering a comprehensive understanding of algorithms' capabilities. In particular, cooperative MARL algorithms are predominantly evaluated on benchmarks such as SMAC and GRF, which primarily feature team game scenarios without assessing adequately various aspects of agents' capabilities required in fully cooperative real-world tasks such as multi-robot cooperation and warehouse, resource management, search and rescue, and human-AI cooperation. Moreover, MARL algorithms are mainly evaluated on low dimensional state spaces, and thus their performance on high-dimensional (e.g., image) observations is not well-studied. To fill this gap, this paper highlights the crucial need for expanding systematic evaluation across a wider array of existing benchmarks. To this end, we conduct extensive evaluation and comparisons of well-known MARL algorithms on complex fully cooperative benchmarks, including tasks with images as agents' observations. Interestingly, our analysis shows that many algorithms, hailed as state-of-the-art on SMAC and GRF, may underperform standard MARL baselines on fully cooperative benchmarks. Finally, towards more systematic and better evaluation of cooperative MARL algorithms, we have open-sourced PyMARLzoo+, an extension of the widely used (E)PyMARL libraries, which addresses an open challenge from [TBG++21], facilitating seamless integration and support with all benchmarks of PettingZoo, as well as Overcooked, PressurePlate, Capture Target and Box Pushing.

cs.LG↗
arXiv2022

Automating the resolution of flight conflicts: Deep reinforcement learning in service of air traffic controllers

Dense and complex air traffic scenarios require higher levels of automation than those exhibited by tactical conflict detection and resolution (CD\&R) tools that air traffic controllers (ATCO) use today. However, the air traffic control (ATC) domain, being safety critical, requires AI systems to which operators are comfortable to relinquishing control, guaranteeing operational integrity and automation adoption. Two major factors towards this goal are quality of solutions, and transparency in decision making. This paper proposes using a graph convolutional reinforcement learning method operating in a multiagent setting where each agent (flight) performs a CD\&R task, jointly with other agents. We show that this method can provide high-quality solutions with respect to stakeholders interests (air traffic controllers and airspace users), addressing operational transparency issues.

cs.MA↗
arXiv2016

Supersymmetric geometries of IIA supergravity III

We find that (massive) IIA backgrounds that admit a $G_2\ltimes \mathbb{R}^8$ invariant Killing spinor must exhibit a null Killing vector field which leaves the Killing spinor invariant and that the rotation of the Killing vector field satisfies a certain $\mathfrak{g}_2$ instanton condition. This result together with those in [4] and [5] complete the classification of geometries of all (massive) IIA backgrounds that preserve one supersymmetry. We also explore the geometry of a class of backgrounds which admit a $G_2\ltimes \mathbb{R}^8$ invariant Killing spinor and where in addition an appropriate 1-form bilinear vanishes. In all cases, we express the fluxes of the theory in terms of the geometry.

hep-th↗
arXiv2015

Supersymmetric geometries of IIA supergravity II

We solve the Killing spinor equations of standard and massive IIA supergravities for a Killing spinor whose isotropy subgroup in Spin(9, 1) is SU(4) and identify the geometry of the spacetime. We demonstrate that the Killing spinor equations impose some mild constraints on the geometry of the spacetime which include the existence of a time-like Killing vector field which leaves the fields and the Killing spinor invariant.

hep-th↗
arXiv2014

Supersymmetric geometries of IIA supergravity I

IIA supergravity backgrounds preserving one supersymmetry locally admit four types of Killing spinors distinguished by the orbits of $Spin(9,1)$ on the space of spinors. We solve the Killing spinor equations of IIA supergravity with and without cosmological constant for Killing spinors representing two of these orbits, with isotropy groups $Spin(7)$ and $Spin(7)\ltimes\mathbb{R}^8$. In both cases, we identify the geometry of spacetime and express the fluxes in terms of the geometry. We find that the geometric constraints of backgrounds with a $Spin(7)\ltimes\mathbb{R}^8$ invariant Killing spinor are identical to those found for heterotic backgrounds preserving one supersymmetry.

hep-th↗
arXiv2013

Semi-global symplectic invariants of the Euler top

We compute the semi-global symplectic invariants near the hyperbolic equilibrium points of the Euler top. The Birkhoff normal form at the hyperbolic point is computed using Lie series. The actions near the hyperbolic point are found using Frobenius expansion of its Picard-Fuchs equation. We show that the Birkhoff normal form can also be found by inverting the regular solution of the Picard-Fuchs equation. Composition of the singular action integral with the Birkhoff normal form gives the semi-global symplectic invariant. Finally, we discuss the convergence of these invariants and show that in a neighbourhood of the separatrix the pendulum is not symplectically equivalent to any Euler top.

math.SG↗
arXiv2012

(1,0) superconformal theories in six dimensions and Killing spinor equations

We solve the Killing spinor equations of 6-dimensional (1,0) superconformal theories in all cases. In particular, we derive the conditions on the fields imposed by the Killing spinor equations and demonstrate that these depend on the isotropy group of the Killing spinors. We focus on the models proposed by Samtleben et al in \cite{ssw} and find that there are solutions preserving 1,2, 4 and 8 supersymmetries. We also explore the solutions which preserve 4 supersymmetries and find that many models admit string and 3-brane solitons as expected from the M-brane intersection rules. The string solitons are smooth regulated by the moduli of instanton configurations.

hep-th↗
arXiv2012

Vanishing theorems on (l|k)-strong Kaehler manifolds with torsion

We derive sufficient conditions for the vanishing of plurigenera, $p_m(J), m>0$, on compact (l|k)-strong, $ω^l\wedge \partial\bar\partial ω^k=0$, Kaehler manifolds with torsion. In particular, we show that the plurigenera of compact (l|k)-strong manifolds, k<n-1, for which the holonomy of the unique Hermitian connection with skew-symmetric torsion is contained in SU(n) vanish. As a consequence all generalized k-Gauduchon manifolds with holonomy of the Hermitian connection with skew-symmetric torsion contained in SU(n) do not admit holomorphic (n,0) forms. Furthermore we show that all conformally balanced, (l|k)-strong Kaehler manifolds with torsion, k<n-1, are Kähler. We also give several compact examples of (l|k)-strong Kaehler and Calabi-Yau manifolds with torsion.

math.DG↗
arXiv2012

Topology and geometry of 6-dimensional (1,0) supergravity black hole horizons

We show that the supersymmetric near horizon black hole geometries of 6-dimensional supergravity coupled to any number of scalar and tensor multiplets are either locally $AdS_3\times Σ^3$, where Σ^3 is a homology 3-sphere, or $\bR^{1,1}\times {\cal S}^4$, where ${\cal S}^4$ is a 4-manifold whose geometry depends on the hypermultiplet scalars. In both cases, we find that the tensorini multiplet scalars are constant and the associated 3-form field strengths vanish. We also demonstrate that the $AdS_3\times Σ^3$ horizons preserve 2, 4 and 8 supersymmetries. For horizons with 4 supersymmetries, Σ^3 is in addition a non-trivial circle fibration over a topological 2-sphere. The near horizon geometries preserving 8 supersymmetries are locally isometric to either $AdS_3\times S^3$ or $\bR^{1,1}\times T^4$. Moreover, we show that the $\bR^{1,1}\times {\cal S}$ horizons preserve 1, 2 and 4 supersymmetries and the geometry of ${\cal S}$ is Riemann, Kähler and hyper-Kähler, respectively.

hep-th↗
arXiv2011

Spinorial geometry and Killing spinor equations of 6-D supergravity

We solve the Killing spinor equations of 6-dimensional (1,0)-supergravity coupled to any number of tensor, vector and scalar multiplets in all cases. The isotropy groups of Killing spinors are $Sp(1)\cdot Sp(1)\ltimes \bH (1)$, $U(1)\cdot Sp(1)\ltimes \bH (2)$, $Sp(1)\ltimes \bH (3,4)$, $Sp(1) (2)$, $U(1) (4)$ and $\{1\} (8)$, where in parenthesis is the number of supersymmetries preserved in each case. If the isotropy group is non-compact, the spacetime admits a parallel null 1-form with respect to a connection with torsion the 3-form field strength of the gravitational multiplet. The associated vector field is Killing and the 3-form is determined in terms of the geometry of spacetime. The $Sp(1)\ltimes \bH$ case admits a descendant solution preserving 3 out of 4 supersymmetries due to the hyperini Killing spinor equation. If the isotropy group is compact, the spacetime admits a natural frame constructed from 1-form spinor bi-linears. In the $Sp(1)$ and U(1) cases, the spacetime admits 3 and 4 parallel 1-forms with respect to the connection with torsion, respectively. The associated vector fields are Killing and under some additional restrictions the spacetime is a principal bundle with fibre a Lorentzian Lie group. The conditions imposed by the Killing spinor equations on all other fields are also determined.

hep-th↗
arXiv2011

The Galactic Census of High- and Medium-mass Protostars. I. Catalogues and First Results from Mopra HCO+ Maps

The Census of High- and Medium-mass Protostars (CHaMP) is the first large-scale, unbiased, uniform mapping survey at sub-parsec scale resolution of 90 GHz line emission from massive molecular clumps in the Milky Way. We present the first Mopra (ATNF) maps of the CHaMP survey region (300°>l>280°) in the HCO+ J=1-0 line, which is usually thought to trace gas at densities up to 10^11 m-3. In this paper we introduce the survey and its strategy, describe the observational and data reduction procedures, and give a complete catalogue of moment maps of the HCO+ J=1-0 emission from the ensemble of 301 massive molecular clumps. From these maps we also derive the physical parameters of the clumps, using standard molecular spectral-line analysis techniques. This analysis yields the following range of properties: integrated line intensity 1-30 K km s-1, peak line brightness 1-7 K, linewidth 1-10 km s-1, integrated line luminosity 0.5-200 K km s-1 pc^2, FWHM size 0.2-2.5 pc, mean projected axial ratio 2, optical depth 0.08-2, total surface density 30-3000 M{\sun} pc-2, number density 0.2-30 x 10^9 m-3, mass 15-8000 M{\sun}, virial parameter 1-55, and total gas pressure 0.3-700 pPa. We find that the CHaMP clumps do not obey a Larson-type size-linewidth relation. Among the clumps, there exists a large population of subthermally excited, weakly-emitting (but easily detectable) dense molecular clumps, confirming the prediction of Narayanan et al. (2008). These weakly-emitting clumps comprise 95% of all massive clumps by number, and 87% of the molecular mass, in this portion of the Galaxy; their properties are distinct from the brighter massive star-forming regions that are more typically studied. If the clumps evolve by slow contraction, the 95% of fainter clumps may represent a long-lived stage of pressure-confined, gravitationally stable massive clump evolution, while the CHaMP ... (abridged)

astro-ph.GA↗
arXiv2010

Covariantly constant forms on torsionful geometries from world-sheet and spacetime perspectives

The symmetries of two-dimensional supersymmetric sigma models on target spaces with covariantly constant forms associated to special holonomy groups are analysed. It is shown that each pair of such forms gives rise to a new one, called a Nijenhuis form, and that there may be further reductions of the structure group. In many cases of interest there are also covariantly constant one-forms which also give rise to symmetries. These geometries are of interest in the context of heterotic supergravity solutions and the associated reductions are studied from a spacetime point of view via the Killing spinor equations.

hep-th↗
arXiv2009

Heterotic supersymmetric backgrounds with compact holonomy revisited

We simplify the classification of supersymmetric solutions with compact holonomy of the Killing spinor equations of heterotic supergravity using the field equations and the additional assumption that the 3-form flux is closed. We determine all the fractions of supersymmetry that the solutions preserve and find that there is a restriction on the number of supersymmetries which depends on the isometry group of the background. We examine the geometry of spacetime in all cases. We find that the supersymmetric solutions of heterotic supergravity are associated with a large number of geometric structures which include 7-dimensional manifolds with $G_2$ structure, 6-dimensional complex and almost complex manifolds, and 4-dimensional hyper-Kähler, Kähler and anti-self-dual Weyl manifolds.

hep-th↗
arXiv2009

New half supersymmetric solutions of the heterotic string

We describe all supersymmetric solutions of the heterotic string which preserve 8 supersymmetries and show that are distinguished by the holonomy, ${\rm hol}(\hat\nabla)$, of the connection, $\hat\nabla$, with skew-symmetric torsion. The ${\rm hol}(\hat\nabla) \subseteq SU(2)$ solutions are principal bundles over a 4-dimensional hyper-Kähler manifold equipped with a anti-self-dual connection and fibre group $G$ which has Lie algebra, ${\mathfrak Lie} (G)=\bR^{5,1}$, $\mathfrak{sl}(2,\bR)\oplus \mathfrak{su}(2)$ or $\mathfrak{cw}_6$. Some of the solutions have the interpretation as 5-branes wrapped on $G$ with transverse space any hyper-Kähler 4-dimensional manifold. We construct new solutions for ${\mathfrak Lie} (G)=\mathfrak{sl}(2,\bR)\oplus \mathfrak{su}(2)$ and show that are characterized by 3 integers and have continuous moduli. There is also a smooth family in this class with one asymptotic region and the dilaton is bounded everywhere on the spacetime. We also demonstrate that the worldvolume theory of the backgrounds with holonomy SU(2) can be understood in terms of gauged WZW models for which the gauge fields are composite. The ${\rm hol}(\hat\nabla) \subseteq\bR^8$ solutions are superpositions of fundamental strings and pp-waves in flat space, which may also include a null rotation. The ${\rm hol}(\hat\nabla)=\{1\}$ heterotic string backgrounds which preserve 8 supersymmetries are Lorentzian group manifolds.

hep-th↗
arXiv2007

Supersymmetric heterotic string backgrounds

We present the main features of the solution of the gravitino and dilatino Killing spinor equations derived in hep-th/0510176 and hep-th/0703143 which have led to the classification of geometric types of all type I backgrounds. We then apply these results to the supersymmetric backgrounds of the heterotic string. In particular, we solve the gaugino Killing spinor equation together with the other two Killing spinor equations of the theory. We also use our results to classify all supersymmetry conditions of ten-dimensional gauge theory.

hep-th↗