arXiv subjects

Yunfeng Hu

Publications and source records attributed to Yunfeng Hu.

8 recordsLinked to original sources

arXiv2026

PANDA: A Matrix-Free Differentiable NMPC Solver via Proximal Averaged Quasi-Newton with Adaptive Linesearch Algorithm

Differentiable nonlinear model predictive control (NMPC) provides a principled way to embed optimal control structure into end-to-end learning paradigms, but its practical use is often limited by the computational and memory costs of both forward optimization and backward sensitivity propagation. This brief proposes PANDA, a matrix-free solver for differentiable NMPC. In the forward pass, PANDA combines proximal-gradient iterations with quasi-Newton acceleration and introduces an adaptive stepsize enlargement mechanism to mitigate the conservativeness of monotone stepsize reduction. The resulting stepsize behavior and its effect on local convergence are theoretically analyzed. In the backward pass, PANDA performs implicit differentiation from the residual equation and computes adjoint sensitivities using Krylov-subspace iterative methods together with automatic-differentiation-based Matrix-Vector product operators, thereby avoiding explicit Hessian and Jacobian construction. The method is evaluated on a nonconvex trailer NMPC problem embedded in an imitation learning task. The results show that PANDA achieves much faster forward and backward computation and lower memory overhead than representative differentiable optimization solvers, while maintaining effective imitation learning performance.

eess.SY↗

arXiv2026

Open-Ended Instruction Realization with LLM-Enabled Multi-Planner Scheduling in Autonomous Vehicles

Most Human-Machine Interaction (HMI) research overlooks the maneuvering needs of passengers in autonomous driving (AD). Natural language offers an intuitive interface, yet translating passenger open-ended instructions into control signals, without sacrificing interpretability and traceability, remains a challenge. This study proposes an instruction-realization framework that leverages a large language model (LLM) to interpret instructions, generates executable scripts that schedule multiple model predictive control (MPC)-based motion planners based on real-time feedback, and converts planned trajectories into control signals. This scheduling-centric design decouples semantic reasoning from vehicle control at different timescales, establishing a transparent, traceable decision-making chain from high-level instructions to low-level actions. Due to the absence of high-fidelity evaluation tools, this study introduces a benchmark for open-ended instruction realization in a closed-loop setting. Comprehensive experiments reveal that the framework significantly improves task-completion rates over instruction-realization baselines, reduces LLM query costs, achieves safety and compliance on par with specialized AD approaches, and exhibits considerable tolerance to LLM inference latency. For more qualitative illustrations and a clearer understanding.

cs.RO↗

arXiv2025

A Gauss-Newton-Induced Structure-Exploiting Algorithm for Differentiable Optimal Control

Differentiable optimal control, particularly differentiable nonlinear model predictive control (NMPC), provides a powerful framework that enjoys the complementary benefits of machine learning and control theory. A key enabler of differentiable optimal control is the computation of derivatives of the optimal trajectory with respect to problem parameters, i.e., trajectory derivatives. Previous works compute trajectory derivatives by solving a differential Karush-Kuhn-Tucker (KKT) system, and achieve this efficiently by constructing an equivalent auxiliary system. However, we find that directly exploiting the matrix structures in the differential KKT system yields significant computation speed improvements. Motivated by this insight, we propose FastDOC, which applies a Gauss-Newton approximation of Hessian and takes advantage of the resulting block-sparsity and positive semidefinite properties of the matrices involved. These structural properties enable us to accelerate the computationally expensive matrix factorization steps, resulting in a factor-of-two speedup in theoretical computational complexity, and in a synthetic benchmark FastDOC achieves up to a 180% time reduction compared to the baseline method. Finally, we validate the method on an imitation learning task for human-like autonomous driving, where the results demonstrate the effectiveness of the proposed FastDOC in practical applications.

eess.SY↗

arXiv2025

From Failures to Fixes: LLM-Driven Scenario Repair for Self-Evolving Autonomous Driving

Ensuring robust and generalizable autonomous driving requires not only broad scenario coverage but also efficient repair of failure cases, particularly those related to challenging and safety-critical scenarios. However, existing scenario generation and selection methods often lack adaptivity and semantic relevance, limiting their impact on performance improvement. In this paper, we propose \textbf{SERA}, an LLM-powered framework that enables autonomous driving systems to self-evolve by repairing failure cases through targeted scenario recommendation. By analyzing performance logs, SERA identifies failure patterns and dynamically retrieves semantically aligned scenarios from a structured bank. An LLM-based reflection mechanism further refines these recommendations to maximize relevance and diversity. The selected scenarios are used for few-shot fine-tuning, enabling targeted adaptation with minimal data. Experiments on the benchmark show that SERA consistently improves key metrics across multiple autonomous driving baselines, demonstrating its effectiveness and generalizability under safety-critical conditions.

cs.CV↗

arXiv2023

Adversarial Driving Behavior Generation Incorporating Human Risk Cognition for Autonomous Vehicle Evaluation

Autonomous vehicle (AV) evaluation has been the subject of increased interest in recent years both in industry and in academia. This paper focuses on the development of a novel framework for generating adversarial driving behavior of background vehicle interfering against the AV to expose effective and rational risky events. Specifically, the adversarial behavior is learned by a reinforcement learning (RL) approach incorporated with the cumulative prospect theory (CPT) which allows representation of human risk cognition. Then, the extended version of deep deterministic policy gradient (DDPG) technique is proposed for training the adversarial policy while ensuring training stability as the CPT action-value function is leveraged. A comparative case study regarding the cut-in scenario is conducted on a high fidelity Hardware-in-the-Loop (HiL) platform and the results demonstrate the adversarial effectiveness to infer the weakness of the tested AV.

cs.AI↗

arXiv2019

From classical to modern opinion dynamics

In this age of Facebook, Instagram and Twitter, there is rapidly growing interest in understanding network-enabled opinion dynamics in large groups of autonomous agents. The phenomena of opinion polarization, the spread of propaganda and fake news, and the manipulation of sentiment are of interest to large numbers of organizations and people, some of whom are resource rich. Whether it is the more nefarious players such as foreign governments that are attempting to sway elections or large corporations that are trying to bend sentiment -- often quite surreptitiously, or it is more open and above board, like researchers that want to spread the news of some finding or some business interest that wants to make a large group of people aware of genuinely helpful innovations that they are marketing, what is at stake is often significant. In this paper we review many of the classical, and some of the new, social interaction models aimed at understanding opinion dynamics. While the first papers studying opinion dynamics appeared over 60 years ago, there is still a great deal of room for innovation and exploration. We believe that the political climate and the extraordinary (even unprecedented) events in the sphere of politics in the last few years will inspire new interest and new ideas. It is our aim to help those interested researchers understand what has already been explored in a significant portion of the field of opinion dynamics. We believe that in doing this, it will become clear that there is still much to be done.

physics.soc-ph↗

arXiv2018

Median Shapes

We introduce and begin to explore the mean and median of finite sets of shapes represented as integral currents. The median can be computed efficiently in practice, and we focus most of our theoretical and computational attention on medians. We consider questions on the existence and regularity of medians. While the median might not exist in all cases, we show that a mass-regularized median is guaranteed to exist. When the input shapes are modeled by integral currents with shared boundaries in codimension $1$, we show that the median is guaranteed to exist, and is contained in the \emph{envelope} of the input currents. On the other hand, we show that medians can be \emph{wild} in this setting, and smooth inputs can generate non-smooth medians. For higher codimensions, we show that \emph{books} are minimizing for a finite set of $1$-currents in $\Bbb{R}^3$ with shared boundaries. As part of this proof, we present a new result in graph theory---that \emph{cozy} graphs are \emph{comfortable}---which should be of independent interest. Further, we show that regular points on the median have book-like tangent cones in this case. From the point of view of computation, we study the median shape in the settings of a finite simplicial complex. When the input shapes are represented by chains of the simplicial complex, we show that the problem of finding the median shape can be formulated as an integer linear program. This optimization problem can be solved as a linear program in practice, thus allowing one to compute median shapes efficiently. We provide open source code implementing our methods, which could also be used by anyone to experiment with ideas of their own. The software could be accessed at https://github.com/tbtraltaa/medianshape.

math.DG↗

arXiv2017

On the Measure of the Midpoints of the Cantor Set in $\mathbb{R}$

In this paper, we are going to discuss the following problem: Let $T$ be a fixed set in $\mathbb{R}^n$. And let $S$ and $B$ he two subsets in $\mathbb{R}^n$ such that for any $x$ in $S$, there exists an $r$ such that $x+ r T$ is a subset of $B$. How small can be $B$ be if we know the size of $S$? Stein proved that for $n$ is greater than or equal to 3 and $T$ is a sphere centered at origin, then $S$ has positive measure implies $B$ has positive measure using spherical maximal operator. Later, Bourgain and Marstrand proved the similar result for $n =2$. And we found an example for why the result fails for $n=1$.

math.MG↗