arXiv subjects

Mingkai Liu

Publications and source records attributed to Mingkai Liu.

16 recordsLinked to original sources

arXiv2026

Semantic-Guided Progressive Object Removal with Gaussian Splatting

Removing unwanted objects from reconstructed 3D scenes is an important task in computer vision, supporting applications in AR/VR, robotics, and digital content creation. Existing methods typically complete the entire masked region in a single step and without effectively utilizing semantic information from other views, leading to difficulties in handling complex geometric details and textures. In this work, we propose a novel framework that integrates Semantic-guided Block Matching (SBM) and Region-Wise Progressive Refinement (RPR) for high-quality 3D object removal. First, we leverage DINOv2 to encode semantic guidance from multi-view observations, and the best match tokens are decoded to complete missing regions in the target view while maintaining cross-view consistency. Second, we introduce a RPR strategy that segments the target mask into multiple subregions and selectively refines those with poor visual quality. Our method is built upon Gaussian Splatting, ensuring high-fidelity scene reconstruction with efficient computation. Experimental results demonstrate that our approach outperforms existing Gaussian-based methods in terms of perceptual quality and coherence in 3D object removal.

cs.RO↗

arXiv2026

SPACE: Swarm Pheromone Fields for Adaptive Collision-Aware Exploration

Massive robot swarms can explore unknown environments quickly, but adding robots eventually stops helping. Doorways and dense traffic create congestion, increasing inter-robot contacts and reducing the value of each additional robot. We study this safety-efficiency tradeoff for ground swarms of tens to hundreds of robots. We present SPACE, Swarm Pheromone Fields for Adaptive Collision-Aware Exploration. Inspired by ant foraging, SPACE maintains a shared environmental field with an attractive frontier pheromone, a repellent explore pheromone, and a fast robot-density field. Coordination is decentralized and mediated through this field. We evaluate SPACE on real building floorplans, namely sixteen home layouts from the HouseExpo dataset and eight campus floors from the KTH dataset, with swarms of up to two hundred and fifty-six robots. SPACE lies on the empirical Pareto frontier. It attains the lowest inter-robot contact rate at every congested swarm size, four to seventeen times fewer than a greedy nearest-frontier planner, while keeping coverage time within about two percent of that near time-optimal planner. The results indicate that, at this scale, coordination mainly improves safety rather than coverage time.

cs.RO↗

arXiv2026

ACEsplat: Accelerated 3D Gaussian Scene Regression via RGB and Poses Only

Per-scene 3D Gaussian Splatting (3DGS) enables high-fidelity rendering, but practical robotic and AR scene capture pipelines often depend on external geometric initialization (e.g., SfM point clouds or depth estimates), which can be slow and brittle in on-site deployment. We present ACEsplat, a fast per-scene optimization framework that reconstructs 3D Gaussian representations from RGB images and camera poses only, without requiring external 3D priors (e.g., precomputed SfM models or supervised depth maps). ACEsplat uses a two-stage pipeline: (1) a self-supervised scene coordinate regression (SCR) module builds an internal geometry prior within 4--5 minutes; (2) SCR features and coordinate priors are fused by a lightweight Gaussian initialization head, followed by per-scene 3DGS optimization. On static-view rendering, ACEsplat achieves 29.11 dB PSNR on Wayspots with real-time SLAM poses and 33.20 dB on Cambridge Landmarks with SfM-refined poses. On RealEstate10K sparse-view novel view synthesis, it achieves competitive image fidelity under a challenging 2-view setting. ACEsplat completes scene-specific SCR mapping and 3DGS reconstruction within 15--25 minutes on a single GPU, making it a practical RGB+pose-only solution for rapid scene setup in robotics and mixed-reality applications.

cs.RO↗

arXiv2026

All-optical control of nonlinear emission from resonant metasurfaces

Nonlinear optics underpins a broad range of photonic technologies, from classical and quantum light sources to emerging nonlinear photonic neural networks. Yet, conventional nonlinear optical devices exhibit static functionality: their transfer characteristics and emission profiles are dictated by the intrinsic nonlinear process and locked by fabrication, limiting adaptability. Here, we introduce an ultra-thin metasurface platform that enables dynamic reconfiguration of nonlinear functionality in a contact-less fashion. By leveraging all-optical control of the optical torque exerted on liquid crystal molecules infiltrating a resonant metasurface, we achieve tunable polynomial nonlinear transfer functions based on third-harmonic generation process. This mechanism further allows real-time modulation of nonlinear weighting across different diffraction orders, revealing a previously unexplored interplay between mode structure and nonlinear emission. Our approach opens up a pathway toward field-programmable nonlinear photonic systems, offering unprecedented flexibility for reconfigurable nonlinear signal processing and adaptive photonic computing.

physics.optics↗

arXiv2026

MotiMem: Motion-Aware Approximate Memory for Energy-Efficient Neural Perception in Autonomous Vehicles

High-resolution sensors are critical for robust autonomous perception but impose a severe memory wall on battery-constrained electric vehicles. In these systems, data movement energy often outweighs computation. Traditional image compression is ill-suited as it is semantically blind and optimizes for storage rather than bus switching activity. We propose MotiMem, a hardware-software co-designed interface. Exploiting temporal coherence,MotiMem uses lightweight 2D Motion Propagation to dynamically identify Regions of Interest (RoI). Complementing this, a Hybrid Sparsity-Aware Coding scheme leverages adaptive inversion and truncation to induce bitlevel sparsity. Extensive experiments across nuScenes, Waymo, and KITTI with 16 detection models demonstrate that MotiMem reduces memory-interface dynamic energy by approximately 43 percent while retaining approximately 93 percent of the object detection accuracy, establishing a new Pareto frontier significantly superior to standard codecs like JPEG and WebP.

cs.CV↗

arXiv2026

Hybrid Cross-Device Localization via Neural Metric Learning and Feature Fusion

We present a hybrid cross-device localization pipeline developed for the CroCoDL 2025 Challenge. Our approach integrates a shared retrieval encoder and two complementary localization branches: a classical geometric branch using feature fusion and PnP, and a neural feed-forward branch (MapAnything) for metric localization conditioned on geometric inputs. A neural-guided candidate pruning strategy further filters unreliable map frames based on translation consistency, while depth-conditioned localization refines metric scale and translation precision on Spot scenes. These components jointly lead to significant improvements in recall and accuracy across both HYDRO and SUCCU benchmarks. Our method achieved a final score of 92.62 (R@0.5m, 5{\deg}) during the challenge.

cs.CV↗

arXiv2025

MACE: Mixture-of-Experts Accelerated Coordinate Encoding for Large-Scale Scene Localization and Rendering

Efficient localization and high-quality rendering in large-scale scenes remain a significant challenge due to the computational cost involved. While Scene Coordinate Regression (SCR) methods perform well in small-scale localization, they are limited by the capacity of a single network when extended to large-scale scenes. To address these challenges, we propose the Mixed Expert-based Accelerated Coordinate Encoding method (MACE), which enables efficient localization and high-quality rendering in large-scale scenes. Inspired by the remarkable capabilities of MOE in large model domains, we introduce a gating network to implicitly classify and select sub-networks, ensuring that only a single sub-network is activated during each inference. Furtheremore, we present Auxiliary-Loss-Free Load Balancing(ALF-LB) strategy to enhance the localization accuracy on large-scale scene. Our framework provides a significant reduction in costs while maintaining higher precision, offering an efficient solution for large-scale scene applications. Additional experiments on the Cambridge test set demonstrate that our method achieves high-quality rendering results with merely 10 minutes of training.

cs.CV↗

arXiv2025

Mapping at First Sense: A Lightweight Neural Network-Based Indoor Structures Prediction Method for Robot Autonomous Exploration

Autonomous exploration in unknown environments is a critical challenge in robotics, particularly for applications such as indoor navigation, search and rescue, and service robotics. Traditional exploration strategies, such as frontier-based methods, often struggle to efficiently utilize prior knowledge of structural regularities in indoor spaces. To address this limitation, we propose Mapping at First Sense, a lightweight neural network-based approach that predicts unobserved areas in local maps, thereby enhancing exploration efficiency. The core of our method, SenseMapNet, integrates convolutional and transformerbased architectures to infer occluded regions while maintaining computational efficiency for real-time deployment on resourceconstrained robots. Additionally, we introduce SenseMapDataset, a curated dataset constructed from KTH and HouseExpo environments, which facilitates training and evaluation of neural models for indoor exploration. Experimental results demonstrate that SenseMapNet achieves an SSIM (structural similarity) of 0.78, LPIPS (perceptual quality) of 0.68, and an FID (feature distribution alignment) of 239.79, outperforming conventional methods in map reconstruction quality. Compared to traditional frontier-based exploration, our method reduces exploration time by 46.5% (from 2335.56s to 1248.68s) while maintaining a high coverage rate (88%) and achieving a reconstruction accuracy of 88%. The proposed method represents a promising step toward efficient, learning-driven robotic exploration in structured environments.

cs.RO↗

arXiv2025

SenseExpo: Spatial Exploration and Navigation via Scene Estimation from Expeditious Predictive Operators

We present \textbf{SenseExpo}, a lightweight single-robot exploration framework that integrates a compact map prediction network into a frontier-based strategy. SenseExpo addresses two long-standing challenges in classical methods -- high computational overhead and poor environmental generalization. Our prediction network combines Generative Adversarial Networks (GANs), Transformers, and Fast Fourier Convolution (FFC) to achieve a remarkably small footprint of only 709K parameters. Despite its compactness, SenseExpo outperforms U-Net (24.5M) and LaMa (51M) on the KTH dataset, achieving PSNR 9.026 and SSIM 0.718, representing a 38.7\% PSNR gain over LaMa. Cross-domain evaluation further verifies strong generalization with an FID of 161.55 on HouseExpo. In exploration experiments, SenseExpo reaches target coverage 67.9\% faster on KTH and 77.1\% faster on MRPB~1.0 than a MapEx-style global obstacle-prediction baseline under the same simulator; because the methods predict different map semantics, this comparison evaluates planning utility rather than a direct predictor ranking. Implemented as a plug-and-play ROS (Robot Operating System) node, our framework integrates with existing navigation stacks, providing an efficient solution for resource-constrained robotic systems.

cs.CV↗

arXiv2022

Topological membrane devices for terahertz on-chip photonics

Terahertz waves offer a profound platform for next-generation sensing, imaging, and information communications. However, all conventional terahertz components and systems suffer from a bulky design, sensitivity to imperfections, and transmission losses. Here, we propose and experimentally demonstrate on-chip integration and miniaturization of topological devices which may address many existing drawbacks of the terahertz technology. We design and fabricate topological devices based on valley-Hall photonic structures that can be employed for various integrated components of on-chip terahertz systems. More specifically, we demonstrate the valley-locked asymmetric energy flow and mode conversion with topological straight waveguide, multi-port couplers, wave division, and whispering gallery mode resonators. Our devices are based on topological membrane metasurfaces which are of great importance for developing on-chip photonics and bringing many novel features into terahertz devices.

physics.optics↗

arXiv2021

Printed tapered leaky-wave antennas for W-band frequencies

Despite their great potential in communication and sensing applications, printed leaky-wave antennas have rarely been reported at mm-wave frequencies. In this paper, tapered leaky-wave antennas operating at 80 GHz are designed, fabricated and experimentally characterized. While most continuous leaky-wave antennas use subwavelength strips or other comparably small elements, in this work, the surface impedance is discretized very coarsely using only three square patches per period. With this architecture, a wide range of surface reactance can be achieved while maintaining a minimum feature size of the metallic pattern that is feasible for printed circuit fabrication. As the analytical solution for the bandstructure of sinusoidally modulated reactance surfaces is inaccurate for coarse discretization, we find it using full-wave simulation. In order to control side lobes effectively, we use a tapered aperture illumination according to the Taylor one-parameter distribution. A comprehensive experimental demonstration is presented, including near-field and far-field measurements. Therewith, we verify the designed aperture illumination and we reveal the origin of spurious far-field features. Side lobes are effectively suppressed and spurious radiation is reduced to -18 dB compared to the main lobe.

physics.app-ph↗

arXiv2019

Time-varying metasurfaces for broadband spectral camouflage

The possibility of making an object invisible for detectors has become a topic of considerable interest over the past decades. Most of the studies so far focused on reducing the visibility by reshaping the electromagnetic scattering in the spatial domain. In fact, by manipulating the electromagnetic scattering in the time domain, the visibility of an object can also be reduced. Importantly, unlike previous studies on phase-switched screens and time-varying metasurfaces, where the effect is narrow band due to the dispersive resonance, for microwave frequency range, we introduce a broadband switchable metasurface integrated with p-i-n diodes. The reflection phase of the metasurface can be changed by approximately π over a fractional bandwidth of 76%. By modulating the metasurface quasirandomly in the time domain, the incident narrow-band signal is spread into a white-noiselike spectrum upon reflection, creating a spectral camouflage. The broadband feature of the proposed time-varying metasurface can provide practical insight for various applications, including radar stealth and ultrawide-band wireless communication.

physics.app-ph↗

arXiv2018

Extreme Huygens' metasurfaces based on quasi-bound states in the continuum

We introduce the concept and a generic approach to realize Extreme Huygens' Metasurfaces by bridging the concepts of Huygens' conditions and optical bound states in the continuum. This novel paradigm allows creating Huygens' metasurfaces whose quality factors can be tuned over orders of magnitudes, generating extremely dispersive phase modulation. We validate this concept with a proof-of-concept experiment at the near-infrared wavelengths, demonstrating all-dielectric Huygens' metasurfaces with different quality factors. Our study points out a practical route for controlling the radiative decay rate while maintaining the Huygens' condition, complementing existing Huygens' metasurfaces whose bandwidths are relatively broad and complicated to tune. This novel feature can provide new insight for various applications, including optical sensing, dispersion engineering and pulse-shaping, tunable metasurfaces, metadevices with high spectral selectivity, and nonlinear meta-optics.

physics.optics↗

arXiv2018

Time-varying Huygens' meta-devices for parametric waves

Huygens' metasurfaces have demonstrated almost arbitrary control over the shape of a scattered beam, however, its spatial profile is typically fixed at fabrication time. Dynamic reconfiguration of this beam profile with tunable elements remains challenging, due to the need to maintain the Huygens' condition across the tuning range. In this work, we experimentally demonstrate that a time-varying metadevice which performs frequency conversion can steer transmitted or reflected beams in an almost arbitrary manner, with fully dynamic control. Our time-varying Huygens' metadevice is made of both electric and magnetic meta-atoms with independently controlled modulation, and the phase of this modulation is imprinted on the scattered parametric waves, controlling their shapes and directions. We develop a theory which shows how the scattering directionality, phase and conversion efficiency of sidebands can be manipulated almost arbitrarily. We demonstrate novel effects including all-angle beam steering and frequency-multiplexed functionalities at microwave frequencies around 4 GHz, using varactor diodes as tunable elements. We believe that the concept can be extended to other frequency bands, enabling metasurfaces with arbitrary phase pattern that can be dynamically tuned over the complete 2πrange.

physics.app-ph↗

arXiv2018

Asymmetric metasurfaces with high-$Q$ resonances governed by bound states in the continuum

We reveal that metasurfaces created by seemingly different lattices of (dielectric or metallic) meta-atoms with broken in-plane symmetry can support sharp high-$Q$ resonances that originate from the physics of bound states in the continuum. We prove rigorously a direct link between the bound states in the continuum and the Fano resonances, and develop a general theory of such metasurfaces, suggesting the way for smart engineering of resonances for many applications in nanophotonics and meta-optics.

physics.optics↗

arXiv2013

Twists and turns for metamaterials

We propose and verify experimentally a new concept for achieving strong nonlinear coupling between the electromagnetic and elastic properties in metamaterials. This coupling is provided through a novel degree of freedom in metamaterial design: internal rotation within structural elements. Our meta-atoms have high sensitivity to electromagnetic wave power, and the elastic and electromagnetic properties can be independently designed to optimise the response. We demonstrate a rich range of nonlinear phenomena including self-tuning and bistability, and provide a comprehensive experimental demonstration of the predicted effects.

physics.optics↗