arXiv subjects

Guanglin Xu

Publications and source records attributed to Guanglin Xu.

11 recordsLinked to original sources

arXiv2026

DriveStack-VLA: Render-Teacher Alignment for BEV-Based DeepStack Vision-Language-Action Model

Vision-Language-Action driving models convert a pretrained Vision-Language Model into a driving policy, allowing them to use world knowledge and follow language guidances. However, existing VLA driving models still lack driving-oriented spatial intelligence: their policies are mainly grounded on perspective image tokens and language priors, while precise motion planning requires metric geometry, top-down scene structure, and attention to safety-critical perceptual cues. This limitation makes current models vulnerable to weak visual geometry modeling and perceptual coverage in expert demonstrations. In this paper, we present DriveStack-VLA, a framework built upon a large VLM backbone. To strengthen the spatial grounding of VLA driving, we develop dual visual modeling components. We inject a Bird-Eye-View representation into the Large Language Model decoder through a DeepStack-style connection, and propose Render-Teacher Alignment to align the perceptual focus of real images with that of rasterized images. Furthermore, to bridge the gap in multimodal trajectory selection, we introduce a head-based self-critique module that ranks sampled trajectories and conditionally refines the best one. DriveStack-VLA achieves 91.6 PDMS on NAVSIMv1, 91.0 EPDMS on NAVSIMv2 (with the human penalty filter enabled), and a driving score of 79.49 with a success rate of 56.36\% on the closed-loop Bench2Drive. More visualizations are available on our project page: https://anonymous.4open.science/w/drivestack-vla/.

cs.CV↗

arXiv2026

SparseWorld: Enhancing End-to-End Autonomous Driving via World Models with Sparse Scene Representation

Recently, world models have made significant progress in enhancing end-to-end driving systems through both future situation forecasting and improved scene understanding. However, existing driving world models are typically built upon dense scene representations, causing high computational costs and redundant information. In this paper, we present SparseWorld, a lightweight world model that focuses on predicting only the critical layout of the scene, enabling efficient future forecasting for end-to-end driving systems. SparseWorld first performs autoregressive rollout to forecast future map elements and surrounding agents, enabling the model to learn how driving scenarios evolve over time. It then leverages these predicted futures to refine downstream motion prediction and trajectory planning. Specifically, we propose a Sparse Dreamer that anticipates future instances in the latent space through joint temporal and spatial attention. By interacting with predicted future instances, the motion planner captures more accurate motion patterns and generates more informed and safety-aware trajectories. Extensive experiments demonstrate that SparseWorld significantly reduces collision risk and achieves state-of-the-art performance on the open-loop planning metrics of the nuScenes dataset with a collision rate of 0.05\%. Moreover, it substantially outperforms the baseline method in closed-loop planning metrics on the Bench2Drive benchmark. Supplementary material is available at the project page: https://wryzju.github.io/SparseWorld/.

cs.CV↗

arXiv2025

QiMeng-CPU-v2: Automated Superscalar Processor Design by Learning Data Dependencies

Automated processor design, which can significantly reduce human efforts and accelerate design cycles, has received considerable attention. While recent advancements have automatically designed single-cycle processors that execute one instruction per cycle, their performance cannot compete with modern superscalar processors that execute multiple instructions per cycle. Previous methods fail on superscalar processor design because they cannot address inter-instruction data dependencies, leading to inefficient sequential instruction execution. This paper proposes a novel approach to automatically designing superscalar processors using a hardware-friendly model called the Stateful Binary Speculation Diagram (State-BSD). We observe that processor parallelism can be enhanced through on-the-fly inter-instruction dependent data predictors, reusing the processor's internal states to learn the data dependency. To meet the challenge of both hardware-resource limitation and design functional correctness, State-BSD consists of two components: 1) a lightweight state-selector trained by the simulated annealing method to detect the most reusable processor states and store them in a small buffer; and 2) a highly precise state-speculator trained by the BSD expansion method to predict the inter-instruction dependent data using the selected states. It is the first work to achieve the automated superscalar processor design, i.e. QiMeng-CPU-v2, which improves the performance by about $380\times$ than the state-of-the-art automated design and is comparable to human-designed superscalar processors such as ARM Cortex A53.

cs.AR↗

arXiv2024

AGON: Automated Design Framework for Customizing Processors from ISA Documents

Customized processors are attractive solutions for vast domain-specific applications due to their high energy efficiency. However, designing a processor in traditional flows is time-consuming and expensive. To address this, researchers have explored methods including the use of agile development tools like Chisel or SpinalHDL, high-level synthesis (HLS) from programming languages like C or SystemC, and more recently, leveraging large language models (LLMs) to generate hardware description language (HDL) code from natural language descriptions. However, each method has limitations in terms of expressiveness, correctness, and performance, leading to a persistent contradiction between the level of automation and the effectiveness of the design. Overall, how to automatically design highly efficient and practical processors with minimal human effort remains a challenge. In this paper, we propose AGON, a novel framework designed to leverage LLMs for the efficient design of out-of-order (OoO) customized processors with minimal human effort. Central to AGON is the nano-operator function (nOP function) based Intermediate Representation (IR), which bridges high-level descriptions and hardware implementations while decoupling functionality from performance optimization, thereby providing an automatic design framework that is expressive and efficient, has correctness guarantees, and enables PPA (Power, Performance, and Area) optimization. Experimental results show that superior to previous LLM-assisted automatic design flows, AGON facilitates designing a series of customized OoO processors that achieve on average 2.35 $\times$ speedup compared with BOOM, a general-purpose CPU designed by experts, with minimal design effort.

cs.AR↗

arXiv2021

Improved Decision Rule Approximations for Multi-Stage Robust Optimization via Copositive Programming

We study decision rule approximations for generic multi-stage robust linear optimization problems. We consider linear decision rules for the case when the objective coefficients, the recourse matrices, and the right-hand sides are uncertain, and consider quadratic decision rules for the case when only the right-hand sides are uncertain. The resulting optimization problems are NP-hard but amenable to copositive programming reformulations that give rise to tight conservative approximations. We further enhance these approximations through new piecewise decision rule schemes. Finally, we prove that our proposed approximations are tighter than the state-of-the-art schemes and demonstrate their superiority through numerical experiments.

math.OC↗

arXiv2019

Data-Driven Distributionally Robust Appointment Scheduling over Wasserstein Balls

We study a single-server appointment scheduling problem with a fixed sequence of appointments, for which we must determine the arrival time for each appointment. We specifically examine two stochastic models. In the first model, we assume that all appointees show up at the scheduled arrival times yet their service durations are random. In the second model, we assume that appointees have random no-show behaviors and their service durations are random given that they show up at the appointments. In both models, we assume that the probability distribution of the uncertain parameters is unknown but can be partially observed via a set of historical data, which we view as independent samples drawn from the unknown distribution. In view of the distributional ambiguity, we propose a data-driven distributionally robust optimization (DRO) approach to determine an appointment schedule such that the worst-case (i.e., maximum) expectation of the system total cost is minimized. A key feature of this approach is that the optimal value and the set of optimal schedules thus obtained provably converge to those of the true model, i.e., the stochastic appointment scheduling model with regard to the true probability distribution of the uncertain parameters. While our DRO models are computationally intractable in general, we reformulate them to copositive programs, which are amenable to tractable semidefinite programming problems with high-quality approximations. Furthermore, under some mild conditions, we recast these models as polynomial-sized linear programs. Through an extensive numerical study, we demonstrate that our approach yields better out-of-sample performance than two state-of-the-art methods.

math.OC↗

arXiv2019

An Integer Programming Formulation of the Key Management Problem in Wireless Sensor Networks

With the advent of modern communications systems, much attention has been put on developing methods for securely transferring information between constituents of wireless sensor networks. To this effect, we introduce a mathematical programming formulation for the key management problem, which broadly serves as a mechanism for encrypting communications. In particular, an integer programming model of the q-Composite scheme is proposed and utilized to distribute keys among nodes of a network whose topology is known. Numerical experiments demonstrating the effectiveness of the proposed model are conducted using using a well-known optimization solver package. An illustrative example depicting an optimal encryption for a small-scale network is also presented.

cs.CR↗

arXiv2019

Distributionally Robust Optimization with Confidence Bands for Probability Density Functions

Distributionally robust optimization (DRO) has been introduced for solving stochastic programs where the distribution of the random parameters is unknown and must be estimated by samples from that distribution. A key element of DRO is the construction of the ambiguity set, which is a set of distributions that covers the true distribution with a high probability. Assuming that the true distribution has a probability density function, we propose a class of ambiguity sets based on confidence bands of the true density function. The use of the confidence band enables us to take the prior knowledge of the shape of the underlying density function into consideration (e.g., unimodality or monotonicity). Using the confidence band constructed by density estimation techniques as the ambiguity set, we establish the convergence of the optimal value of DRO to that of the stochastic program as the sample size increases. However, the resulting DRO problem is computationally intractable, as it involves functional decision variables as well as infinitely many constraints. To address this challenge, using the duality theory, we reformulate it into a finite-dimensional stochastic program, which is amenable to a stochastic subgradient scheme as a solution method. We compare our approach with existing state-of-the-art DRO methods on the newsvendor problem and the portfolio management problem, and the numerical results showcase the advantage of our approach.

math.OC↗

arXiv2017

A Data-Driven Distributionally Robust Bound on the Expected Optimal Value of Uncertain Mixed 0-1 Linear Programming

This paper studies the expected optimal value of a mixed 0-1 programming problem with uncertain objective coefficients following a joint distribution. We assume that the true distribution is not known exactly, but a set of independent samples can be observed. Using the Wasserstein metric, we construct an ambiguity set centered at the empirical distribution from the observed samples and containing the true distribution with a high statistical guarantee. The problem of interest is to investigate the bound on the expected optimal value over the Wasserstein ambiguity set. Under standard assumptions, we reformulate the problem into a copositive program, which naturally leads to a tractable semidefinite-based approximation. We compare our approach with a moment-based approach from the literature on three applications. Numerical results illustrate the effectiveness of our approach.

math.OC↗

arXiv2017

A Copositive Approach for Two-Stage Adjustable Robust Optimization with Uncertain Right-Hand Sides

We study two-stage adjustable robust linear programming in which the right-hand sides are uncertain and belong to a convex, compact uncertainty set. This problem is NP-hard, and the affine policy is a popular, tractable approximation. We prove that under standard and simple conditions, the two-stage problem can be reformulated as a copositive optimization problem, which in turn leads to a class of tractable, semidefinite-based approximations that are at least as strong as the affine policy. We investigate several examples from the literature demonstrating that our tractable approximations significantly improve the affine policy. In particular, our approach solves exactly in polynomial time a class of instances of increasing size for which the affine policy admits an arbitrarily large gap.

math.OC↗

arXiv2015

Robust Sensitivity Analysis of the Optimal Value of Linear Programming

We propose a framework for sensitivity analysis of linear programs (LPs) in minimization form, allowing for simultaneous perturbations in the objective coefficients and right-hand sides, where the perturbations are modeled in a compact, convex uncertainty set. This framework unifies and extends multiple approaches for LP sensitivity analysis in the literature and has close ties to worst-case linear optimization and two-stage adaptive optimization. We define the minimum (best-case) and maximum (worst-case) LP optimal values, p- and p+, over the uncertainty set, and we discuss issues of finiteness, attainability, and computational complexity. While p- and p+ are difficult to compute in general, we prove that they equal the optimal values of two separate, but related, copositive programs. We then develop tight, tractable conic relaxations to provide lower and upper bounds on p- and p+, respectively. We also develop techniques to assess the quality of the bounds, and we validate our approach computationally on several examples from--and inspired by--the literature. We find that the bounds on p- and p+ are very strong in practice and, in particular, are at least as strong as known results for specific cases from the literature.

math.OC↗