arXiv subjects

Kushal Patel

Publications and source records attributed to Kushal Patel.

12 recordsLinked to original sources

arXiv2026

A Signal Analysis Framework for Unshielded Room-Temperature Magnetocardiography

Room-temperature, unshielded recording of cardiac magnetic signals has remained a significant challenge since the inception of magnetocardiography (MCG). In this work, we present an MCG system based on optically pumped magnetometers (OPMs) designed to operate in ambient magnetic environments and acquire adult human cardiac magnetic fields, without the need for active or passive shielding. The system operates in a gradiometer configuration, achieving background-noise cancellation with a common-mode rejection ratio (CMRR) of 31 dB and a gradient sensitivity of 314 $\mathrm{fT/cm/\sqrt{Hz}}$. MCG signals were acquired sequentially at 16 locations across the anterior thorax, and a comprehensive signal-analysis framework incorporating wavelet multiscale principal component analysis (WMSPCA) filtering and signal quality estimation (SQE) scoring was developed to enhance signal quality. This framework yielded a QRS complex signal-to-noise ratio (SNR) of $28.56 \pm 5.61$ dB across all measurement locations. These results demonstrate the feasibility of performing clinical-grade MCG in unshielded, real-world magnetic environments, with consistent morphological fidelity across the QRS complex and T-wave segments. This work represents a meaningful step toward the practical deployment of OPM-based MCG systems in hospital and point-of-care settings.

cond-mat.stat-mech↗

arXiv2026

DualDiffusion: A Speculative Decoding Strategy for Masked Diffusion Models

Masked Diffusion Models (MDMs) offer a promising alternative to autoregressive language models by enabling parallel token generation and bidirectional context modeling. However, their inference speed is significantly limited by the inability to cache key-value pairs due to bidirectional attention, requiring $O(N^2)$ computations at each generation step. While recent methods like FastDLLM and DkvCache improve inference speed through attention approximations and caching strategies, they achieve speedups at the cost of generation quality. We propose DualDiffusion, a speculative decoding framework for MDMs that combines fast drafter models (using efficient approximations) with slower, more accurate verifier models. By running multiple steps of a lightweight drafter followed by a single verification step, DualDiffusion achieves a superior Pareto frontier between generation steps and accuracy compared to existing approaches. We evaluate our method on MMLU and GSM8K, demonstrating that DualDiffusion maintains high accuracy while reducing the number of generation steps required, effectively pushing the quality-efficiency trade-off curve for masked diffusion language models.

cs.LG↗

arXiv2026

Magnetocardiography measurements using an optically pumped magnetometer under ambient conditions

In this work, we report the development of a rubidium-based single-beam scalar optically pumped magnetometer (OPM) and demonstrate its application in measuring human cardiac magnetic fields in an unshielded environment. The developed magnetometers exhibit a noise floor below 15 pT/sqrt(Hz) in the frequency range of 1 to 35 Hz, with a measurement bandwidth of 100 Hz. When operated in a gradiometric configuration, the noise floor is further reduced to below 3 pT/sqrt(Hz) over the same frequency range. Magnetocardiography (MCG) signals were recorded at five different locations across the thorax. A clear polarity reversal of the QRS complex was observed across these measurement positions, confirming the spatial sensitivity of the system. The proposed system shows strong potential for clinical diagnostics, offering valuable physiological information through non-contact MCG measurements

cond-mat.stat-mech↗

arXiv2024

Asymmetric dynamical localization and precision measurement of BEC micromotion

We employ a Bose Einstein Condensate (BEC) based atom-optic kicked rotor to generate an asymmetrically localized momentum distribution that depends upon initial velocity of the BEC. Asymmetric features are shown to arise from the early-time dynamics induced by the broken parity symmetry and, asymptotically freeze as the dynamical localization stabilizes. The asymmetry in the momentum distribution critically depends upon the initial launch velocity and is sensitive to very small initial velocities ('micromotion') of the BEC. In this work, we also perform a precise measurement of the 'micromotion'. By utilizing the technique of measuring the early-time asymmetry of momentum distribution, we report measurement of micromotion down to (230 \pm 17 , \mu\text{m/s}).

quant-ph↗

arXiv2024

Learning Robot Soccer from Egocentric Vision with Deep Reinforcement Learning

We apply multi-agent deep reinforcement learning (RL) to train end-to-end robot soccer policies with fully onboard computation and sensing via egocentric RGB vision. This setting reflects many challenges of real-world robotics, including active perception, agile full-body control, and long-horizon planning in a dynamic, partially-observable, multi-agent domain. We rely on large-scale, simulation-based data generation to obtain complex behaviors from egocentric vision which can be successfully transferred to physical robots using low-cost sensors. To achieve adequate visual realism, our simulation combines rigid-body physics with learned, realistic rendering via multiple Neural Radiance Fields (NeRFs). We combine teacher-based multi-agent RL and cross-experiment data reuse to enable the discovery of sophisticated soccer strategies. We analyze active-perception behaviors including object tracking and ball seeking that emerge when simply optimizing perception-agnostic soccer play. The agents display equivalent levels of performance and agility as policies with access to privileged, ground-truth state. To our knowledge, this paper constitutes a first demonstration of end-to-end training for multi-agent robot soccer, mapping raw pixel observations to joint-level actions, that can be deployed in the real world. Videos of the game-play and analyses can be seen on our website https://sites.google.com/view/vision-soccer .

cs.RO↗

arXiv2023

Spectroscopy of the $5s5p$ $ ^3 P_0 \rightarrow 5s5d$ $ ^3 D_1 $ transition of strontium using laser cooled atoms

This article presents spectroscopy results of the $5s5p{\;^3}P_0 \rightarrow 5s5d{\;^3}D_1$ transition in all isotopes of laser cooled Sr atoms and the utility of this transition for repumping application. By employing the $5s5p{\;^{3} P_{0}} \rightarrow 5s5d{\;^3}D_1 $ (483 nm) transition in combination with the excitation of $5s5p{\;^3}P_2 \rightarrow 5s6s{\;^3}S_1$ (707 nm) transition, we observe a significant increase ($\sim$ 13 fold) in the steady state number of atoms in the magneto-optic trap (MOT). This enhancement is attributed to the efficient repumping of Sr atoms that have decayed into the dark $5s5p{\;^3}P_2$ state by returning them to the ground state $5s^2{\;^1}S_0$ without any loss into the other states. The absolute transition frequencies were measured with an absolute accuracy of 30 MHz. To support our measurements, we performed Fock-space relativistic coupled-cluster calculations of the relevant parameters in Sr. To further increase the accuracy of the calculated properties, corrections from the Breit, QED and perturbative triples were also included. The calculated branching ratio for the repumping state confirms the significantly increased population in the ${^3}P_1$ state. Thereby, leading to an increase of population of atoms trapped due to the enhanced repumping. Our calculated hyperfine-splitting energies are in excellent agreement with the measured values. Moreover, our calculated isotope shifts in the transition frequencies are in good agreement with our measured values.

physics.atom-ph↗

arXiv2023

Comparative Analysis of Phase Noise for different configurations of Bragg lattice for an Atomic Gravimeter with Bose-Einstein Condensate

We perform a comparative study of the phase noise induced in the lasers used for Bragg diffraction in a Bose-Einstein condensate-based quantum gravimeter where the Bragg beams are generated using two different configurations. In one of the configurations, the Bragg beams that form the moving optical lattice are generated using two different acousto-optic modulators. In the second configuration, the Bragg beams are generated using a single acousto-optic modulator carrying two phase-locked frequencies. The second configuration shows a suppression of phase noise by a factor of 4.7 times in the frequency band upto 10 $kHz$, the primary source of noise, which is the background acoustic noise picked up by optical components and the optical table. We report a sensitivity of 99.7 $\mu Gal/\sqrt Hz$ for an interferometric time of 10 $ms$.

quant-ph↗

arXiv2023

Electromagnetically Induced Transparency (EIT) aided cooling of strontium atoms

The presence of ultra-narrow inter-combination spectroscopic lines in alkaline earth elements places them as promising candidates for optical atomic clocks, quantum computation, and for probing fundmental physics. Doppler cooling of these atoms is typically achieved through two subsequent stages: the initial cooling is on the 1s0-1p1 transition followed by cooling using the narrow-line 1s0-3p1 transition. However, due to significantly lower linewidth of the second stage cooling transition, efficient transfer of atoms into the second stage becomes technically challenging. The velocity distribution of the atoms after the first stage of cooling is too broad for atoms to be captured efficiently in the second stage cooling. As a result, the capture efficiency of atoms into the second stage Magneto-Optical Trap is low, even if the linewidth of the second stage cooling laser is artificially broadened.

physics.atom-ph↗

arXiv2023

Learning Agile Soccer Skills for a Bipedal Robot with Deep Reinforcement Learning

We investigate whether Deep Reinforcement Learning (Deep RL) is able to synthesize sophisticated and safe movement skills for a low-cost, miniature humanoid robot that can be composed into complex behavioral strategies in dynamic environments. We used Deep RL to train a humanoid robot with 20 actuated joints to play a simplified one-versus-one (1v1) soccer game. The resulting agent exhibits robust and dynamic movement skills such as rapid fall recovery, walking, turning, kicking and more; and it transitions between them in a smooth, stable, and efficient manner. The agent's locomotion and tactical behavior adapts to specific game contexts in a way that would be impractical to manually design. The agent also developed a basic strategic understanding of the game, and learned, for instance, to anticipate ball movements and to block opponent shots. Our agent was trained in simulation and transferred to real robots zero-shot. We found that a combination of sufficiently high-frequency control, targeted dynamics randomization, and perturbations during training in simulation enabled good-quality transfer. Although the robots are inherently fragile, basic regularization of the behavior during training led the robots to learn safe and effective movements while still performing in a dynamic and agile way -- well beyond what is intuitively expected from the robot. Indeed, in experiments, they walked 181% faster, turned 302% faster, took 63% less time to get up, and kicked a ball 34% faster than a scripted baseline, while efficiently combining the skills to achieve the longer term objectives.

cs.RO↗

arXiv2022

Control of dynamical localization in atom-optics kicked rotor

Atom-optics kicked rotor represents an experimentally realizable version of the paradigmatic quantum kicked rotor system. After a short initial diffusive phase the cloud settles down to a stationary state due to the onset of dynamical localization. In this work we realise an enhancement of localization by modification of the kick sequence. We experimentally implement the modification to this system in which the sign of the kick sequence is flipped by allowing for a free evolution of the wavepackets for half the Talbot time after every $M$ kicks. Depending on the value of $M$, this modified system displays a combination of enhanced diffusion followed by asymptotic localization. This is explained as resulting from two competing processes -- localization induced by standard kicked rotor type kicks, and diffusion induced by half Talbot time evolution. The evolving states display a localized but non-exponential wave function profiles. This provides another route to quantum control in kicked rotor class of systems. The numerical simulations agree well with the experimental results.

quant-ph↗

arXiv2019

Study of loss dynamics of strontium in a magneto-optical trap

Collisions with background atoms are known to induce a significant shift in the frequency of state-of-the-art optical atomic clocks and contribute to state decoherence in cold atom experiments. The effects of these collisions can be quantified by measuring their cross sections. We experimentally measured the collision cross section between $^{88}$Sr$-$N$_{2}$ in a Magneto-Optical Trap (MOT). The measurement was carried out by monitoring the atom number loss rate as a function of background pressure of N$_{2}$ and the cross section thus obtained was 8.1(4)$\times 10^{-18}$ m$^{2}$. The measured collision cross section has been utilized for the determination of C$_{6}$ coefficient of the ground state (${^1S}_0$) of $^{88}$Sr atom, which can be useful to estimate the relative frequency shift in the clock transition. We also estimate the loss rate induced by the combined effect of the decay of atoms in the long-lived ${^3P}_0$ state and temperature-induced atomic losses from the capture volume of the MOT. We find that the contribution due to the latter is dominant in comparison to the other atomic loss channels and must be included in the studies that rely on the total loss rate measurement.

physics.atom-ph↗

arXiv2018

A simple atomic beam oven with a metal thermal break

We report the design and construction of a simple, easy to machine high-temperature oven for generating an atomic beam in laser cooling experiments. This design eliminates the problem of thermal isolation of the oven region from the rest of the vacuum system without using a glass or ceramic thermal break. This design simplifies the construction and operation of high-temperature ovens for elements having low vapor pressure. We demonstrate the functionality of such a source for Strontium (Sr) atoms. We generate a high flux of Sr atoms for use in laser cooling and trapping experiments. The optimization of the design of the metal thermal break is done using a finite element analysis.

physics.atom-ph↗