arXiv subjects

Howard Huang

Publications and source records attributed to Howard Huang.

13 recordsLinked to original sources

arXiv2026

SAVMap: Structure-Aided Visual Mapping of Large-Scale 2.5D Manhattan Wireframes from Panoramic Video

Precise 3D representations of industrial environments enable tasks such as robot localization and digital twin generation. We propose SAVMap, a method for generating a semantic wireframe map of warehouse shelf and light structures using only a panoramic video camera as the sensor input. Sequences of rectified images with shelf and ceiling-facing views are extracted from a panoramic video captured along the warehouse aisles. Using a semantic segmentation network front end, a set of sparse, semantic structure feature points (e.g., corners of shelf structures, centers of lights) are extracted from each image and tracked across the sequences. By accounting for real-world geometric relationships among the points such as Manhattan grids, a constrained structure-from-motion algorithm yields the 3D points that form a wireframe map. We demonstrate the scalability and accuracy of our proposal in a warehouse with 46 shelving rows, each with faces spanning 55\,m by 7\,m. From an hour of panoramic video content, we create wireframe maps for over 5000 shelf elements across the rows, achieving an aggregate mean absolute error of 4.8\,cm with respect to ground-truth.

cs.CV↗

arXiv2026

GenAI for Systems: Recurring Challenges and Design Principles from Software to Silicon

Generative AI is reshaping how computing systems are designed, optimized, and built, yet research remains fragmented across software, architecture, and chip design communities. This paper takes a cross-stack perspective, examining how generative models are being applied from code generation and distributed runtimes through hardware design space exploration to RTL synthesis, physical layout, and verification. Rather than reviewing each layer in isolation, we analyze how the same structural difficulties and effective responses recur across the stack. Our central finding is one of convergence. Despite the diversity of domains and tools, the field keeps encountering five recurring challenges (the feedback loop crisis, the tacit knowledge problem, trust and validation, co-design across boundaries, and the shift from determinism to dynamism) and keeps arriving at five design principles that independently emerge as effective responses (embracing hybrid approaches, designing for continuous feedback, separating concerns by role, matching methods to problem structure, and building on decades of systems knowledge). We organize these into a challenge--principle map that serves as a diagnostic and design aid, showing which principles have proven effective for which challenges across layers. Through concrete cross-stack examples, we show how systems navigate this map as they mature, and argue that the field needs shared engineering methodology, including common vocabularies, cross-layer benchmarks, and systematic design practices, so that progress compounds across communities rather than being rediscovered in each one. Our analysis covers more than 275 papers spanning eleven application areas across three layers of the computing stack, and distills open research questions that become visible only from a cross-layer vantage point.

cs.SE↗

arXiv2024

TorchTitan: One-stop PyTorch native solution for production ready LLM pre-training

The development of large language models (LLMs) has been instrumental in advancing state-of-the-art natural language processing applications. Training LLMs with billions of parameters and trillions of tokens require sophisticated distributed systems that enable composing and comparing several state-of-the-art techniques in order to efficiently scale across thousands of accelerators. However, existing solutions are complex, scattered across multiple libraries/repositories, lack interoperability, and are cumbersome to maintain. Thus, curating and empirically comparing training recipes require non-trivial engineering effort. This paper introduces TorchTitan, an open-source, PyTorch-native distributed training system that unifies state-of-the-art techniques, streamlining integration and reducing overhead. TorchTitan enables 3D parallelism in a modular manner with elastic scaling, providing comprehensive logging, checkpointing, and debugging tools for production-ready training. It also incorporates hardware-software co-designed solutions, leveraging features like Float8 training and SymmetricMemory. As a flexible test bed, TorchTitan facilitates custom recipe curation and comparison, allowing us to develop optimized training recipes for Llama 3.1 and provide guidance on selecting techniques for maximum efficiency based on our experiences. We thoroughly assess TorchTitan on the Llama 3.1 family of LLMs, spanning 8 billion to 405 billion parameters, and showcase its exceptional performance, modular composability, and elastic scalability. By stacking training optimizations, we demonstrate accelerations of 65.08% with 1D parallelism at the 128-GPU scale (Llama 3.1 8B), an additional 12.59% with 2D parallelism at the 256-GPU scale (Llama 3.1 70B), and an additional 30% with 3D parallelism at the 512-GPU scale (Llama 3.1 405B) on NVIDIA H100 GPUs over optimized baselines.

cs.CL↗

arXiv2021

Domain Adaptation of Networks for Camera Pose Estimation: Learning Camera Pose Estimation Without Pose Labels

One of the key criticisms of deep learning is that large amounts of expensive and difficult-to-acquire training data are required in order to train models with high performance and good generalization capabilities. Focusing on the task of monocular camera pose estimation via scene coordinate regression (SCR), we describe a novel method, Domain Adaptation of Networks for Camera pose Estimation (DANCE), which enables the training of models without access to any labels on the target task. DANCE requires unlabeled images (without known poses, ordering, or scene coordinate labels) and a 3D representation of the space (e.g., a scanned point cloud), both of which can be captured with minimal effort using off-the-shelf commodity hardware. DANCE renders labeled synthetic images from the 3D model, and bridges the inevitable domain gap between synthetic and real images by applying unsupervised image-level domain adaptation techniques (unpaired image-to-image translation). When tested on real images, the SCR model trained with DANCE achieved comparable performance to its fully supervised counterpart (in both cases using PnP-RANSAC for final pose estimation) at a fraction of the cost. Our code and dataset are available at https://github.com/JackLangerman/dance

cs.CV↗

arXiv2021

Feature Learning for Neural-Network-Based Positioning with Channel State Information

Recent channel state information (CSI)-based positioning pipelines rely on deep neural networks (DNNs) in order to learn a mapping from estimated CSI to position. Since real-world communication transceivers suffer from hardware impairments, CSI-based positioning systems typically rely on features that are designed by hand. In this paper, we propose a CSI-based positioning pipeline that directly takes raw CSI measurements and learns features using a structured DNN in order to generate probability maps describing the likelihood of the transmitter being at pre-defined grid points. To further improve the positioning accuracy of moving user equipments, we propose to fuse a time-series of learned CSI features or a time-series of probability maps. To demonstrate the efficacy of our methods, we perform experiments with real-world indoor line-of-sight (LoS) and non-LoS channel measurements. We show that CSI feature learning and time-series fusion can reduce the mean distance error by up to 2.5$\boldsymbol\times$ compared to the state-of-the-art.

cs.IT↗

arXiv2021

Experimental Study on Probabilistic ToA and AoA Joint Localization in Real Indoor Environments

In this paper, we study probabilistic time-of-arrival (ToA) and angle-of-arrival (AoA) joint localization in real indoor environments. To mitigate the effects of multipath propagation, the joint localization algorithm incorporates into the likelihood function Gaussian mixture models (GMM) and the Von Mises-Fisher distribution to model time bias errors and angular uncertainty, respectively. We evaluate the algorithm performance using a proprietary prototype deployed in an indoor factory environment with infrastructure receivers in each of the four corners at the ceiling of a 10 meter by 20 meter section. The field test results show that our joint probabilistic localization algorithm significantly outperforms baselines using only ToA or AoA measurements and achieves 2-D sub-meter accuracy at the 90%-ile. We also numerically demonstrate that the joint localization algorithm is more robust to synchronization errors than the baseline using ToA measurements only.

cs.NI↗

arXiv2020

CSI-Based Multi-Antenna and Multi-Point Indoor Positioning Using Probability Fusion

Channel state information (CSI)-based fingerprinting via neural networks (NNs) is a promising approach to enable accurate indoor and outdoor positioning of user equipments (UEs), even under challenging propagation conditions. In this paper, we propose a positioning pipeline for wireless LAN MIMO-OFDM systems which uses uplink CSI measurements obtained from one or more unsynchronized access points (APs). For each AP receiver, novel features are first extracted from the CSI that are robust to system impairments arising in real-world transceivers. These features are the inputs to a NN that extracts a probability map indicating the likelihood of a UE being at a given grid point. The NN output is then fused across multiple APs to provide a final position estimate. We provide experimental results with real-world indoor measurements under line-of-sight (LoS) and non-LoS propagation conditions for an 80MHz bandwidth IEEE 802.11ac system using a two-antenna transmit UE and two AP receivers each with four antennas. Our approach is shown to achieve centimeter-level median distance error, an order of magnitude improvement over a conventional baseline.

eess.SP↗

arXiv2019

Probabilistic Time of Arrival Localization

In this paper, we take a new approach for time of arrival geo-localization. We show that the main sources of error in metropolitan areas are due to environmental imperfections that bias our solutions, and that we can rely on a probabilistic model to learn and compensate for them. The resulting localization error is validated using measurements from a live LTE cellular network to be less than 10 meters, representing an order-of-magnitude improvement.

cs.LG↗

arXiv2018

Throughput Maximization for Delay-Sensitive Random Access Communication

Future 5G cellular networks supporting ultra-reliable, low-latency communications (URLLC) could employ random access communication to reduce the overhead compared to scheduled access techniques used in 4G networks. We consider a wireless communication system where multiple devices transmit payloads of a given fixed size in a random access fashion over shared radio resources to a common receiver. We allow retransmissions and assume Chase combining at the receiver. The radio resources are partitioned in the time and frequency dimensions, and we determine the optimal partition granularity to maximize throughput, subject to given constraints on latency and outage. In the regime of high and low signal-to-noise ratio (SNR), we derive explicit expressions for the granularity and throughput, first using a Shannon capacity approximation and then using finite block length analysis. Numerical results show that the throughput scaling results are applicable over a range of SNRs. The proposed analytical framework can provide insights for resource allocation strategies in general random access systems and in specific 5G use cases for massive URLLC uplink access.

cs.IT↗

arXiv2016

Wi-Fi Scaling and Performance in Dense Frequency Planned Networks

The capacity scaling of a Wi-Fi network with fixed transmit power is investigated. The 802.11 MAC appears to impose a hard limit on the area capacity by enforcing single transmissions in overlapped cell coverage areas. However due to the difficulty reading the packet header when transmissions either overlap or are faded, stations in different cells may lose synchronization with each other. A large frequency planned network with full propagation impairments of fast and slow fading is simulated in NS-3 with an upgraded receiver model supporting packet capture. Downlink Wi-Fi performance is compared to an equivalent LTE network on the same cell centers. Results show that Wi-Fi downlink throughput grows with densification limited by decreasing SINR and partially operational virtual carrier sense. LTE area capacity grows faster as it is only constrained by SINR.

cs.NI↗

arXiv2015

Wide-area Wireless Communication Challenges for the Internet of Things

Aided by the ubiquitous wireless connectivity, declining communication costs, and the emergence of cloud platforms, the deployment of Internet of Things (IoT) devices and services is accelerating. Most major mobile network operators view machine-to-machine (M2M) communication networks for supporting IoT as a significant source of new revenue. In this paper, we motivate the need for wide-area M2M wireless networks, especially for short data packet communication to support a very large number of IoT devices. We first present a brief overview of current and emerging technologies for supporting wide area M2M, and then using communication theory principles, discuss the fundamental challenges and potential solutions for these networks, highlighting tradeoffs and strategies for random and scheduled access. We conclude with recommendations for how future 5G networks should be designed for efficient wide-area M2M communications.

cs.IT↗

arXiv2010

Linear Precoding in Cooperative MIMO Cellular Networks with Limited Coordination Clusters

In a cooperative multiple-antenna downlink cellular network, maximization of a concave function of user rates is considered. A new linear precoding technique called soft interference nulling (SIN) is proposed, which performs at least as well as zero-forcing (ZF) beamforming. All base stations share channel state information, but each user's message is only routed to those that participate in the user's coordination cluster. SIN precoding is particularly useful when clusters of limited sizes overlap in the network, in which case traditional techniques such as dirty paper coding or ZF do not directly apply. The SIN precoder is computed by solving a sequence of convex optimization problems. SIN under partial network coordination can outperform ZF under full network coordination at moderate SNRs. Under overlapping coordination clusters, SIN precoding achieves considerably higher throughput compared to myopic ZF, especially when the clusters are large.

cs.IT↗

arXiv2009

Efficient Linear Precoding in Downlink Cooperative Cellular Networks with Soft Interference Nulling

A simple line network model is proposed to study the downlink cellular network. Without base station cooperation, the system is interference-limited. The interference limitation is overcome when the base stations are allowed to jointly encode the user signals, but the capacity-achieving dirty paper coding scheme can be too complex for practical implementation. A new linear precoding technique called soft interference nulling (SIN) is proposed, which performs at least as well as zero-forcing (ZF) beamforming under full network coordination. Unlike ZF, SIN allows the possibility of but over-penalizes interference. The SIN precoder is computed by solving a convex optimization problem, and the formulation is extended to multiple-antenna channels. SIN can be applied when only a limited number of base stations cooperate; it is shown that SIN under partial network coordination can outperform full network coordination ZF at moderate SNRs.

cs.IT↗