4.7 Article

Outlier-Robust Estimation: Hardness, Minimally Tuned Algorithms, and Applications

期刊

IEEE TRANSACTIONS ON ROBOTICS
卷 38, 期 1, 页码 281-301

出版社

IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TRO.2021.3094984

关键词

Estimation; Signal processing algorithms; Probabilistic logic; Approximation algorithms; Simultaneous localization and mapping; Particle measurements; Measurement uncertainty; Algorithms; autonomous systems; computational complexity; computer vision; maximum likelihood estimation; resilient perception; robust estimation

类别

资金

  1. ARL DCIST CRA [W911NF-172-0181]
  2. MathWorks
  3. NS CAREER award Certifiable Perception for Autonomous Cyber-Physical Systems,
  4. Lincoln Laboratory's Resilient Perception in Degraded Environments program

向作者/读者索取更多资源

This article introduces two unifying formulations for outlier-robust estimation, generalized maximum consensus and generalized truncated least squares, and proposes a novel algorithm for dynamically separating inliers from outliers. The research shows that these algorithms perform well on robot perception problems, outperforming existing methods even without relying on noise bounds for the inliers.
Nonlinear estimation in robotics and vision is typically plagued with outliers due to wrong data association or incorrect detections from signal processing and machine learning methods. This article introduces two unifying formulations for outlier-robust estimation, generalized maximum consensus ($\textrm{G}$-$\textrm{MC}$) and generalized truncated least squares ($\textrm{G-TLS}$), and investigates fundamental limits, practical algorithms, and applications. Our first contribution is a proof that outlier-robust estimation is inapproximable: In the worst case, it is impossible to (even approximately) find the set of outliers, even with slower-than-polynomial-time algorithms (particularly, algorithms running in quasi-polynomial time). As a second contribution, we review and extend two general-purpose algorithms. The first, adaptive trimming ($\textrm{ADAPT}$), is combinatorial and is suitable for $\textrm{G}$-$\textrm{MC}$; the second, graduated nonconvexity ($\textrm{GNC}$), is based on homotopy methods and is suitable for $\textrm{G-TLS}$. We extend $\textrm{ADAPT}$ and $\textrm{GNC}$ to the case where the user does not have prior knowledge of the inlier-noise statistics (or the statistics may vary over time) and is unable to guess a reasonable threshold to separate inliers from outliers (as the one commonly used in RANdom SAmple Consensus $(\textrm{RANSAC})$. We propose the first minimally tuned algorithms for outlier rejection, which dynamically decide how to separate inliers from outliers. Our third contribution is an evaluation of the proposed algorithms on robot perception problems: mesh registration, image-based object detection (shape alignment), and pose graph optimization. $\textrm{ADAPT}$ and $\textrm{GNC}$ execute in real time, are deterministic, outperform $\textrm{RANSAC}$, and are robust up to 80-90% outliers. Their minimally tuned versions also compare favorably with the state of the art, even though they do not rely on a noise bound for the inliers.

作者

我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。

评论

主要评分

4.7
评分不足

次要评分

新颖性
-
重要性
-
科学严谨性
-
评价这篇论文

推荐

Article Automation & Control Systems

LQG Control and Sensing Co-Design

Vasileios Tzoumas, Luca Carlone, George J. Pappas, Ali Jadbabaie

Summary: The study focuses on the joint design of sensing and control policies, tackling two dual problem instances: sensing-constrained LQG control and minimum-sensing LQG control. While no polynomial time algorithm guarantees a constant approximation factor from the optimal across all problem instances, the authors present the first polynomial time algorithms with per-instance suboptimality guarantees. The research frames LQG codesign as the optimization of approximately supermodular set functions, developing novel algorithms and establishing connections between suboptimality and control-theoretic quantities.

IEEE TRANSACTIONS ON AUTOMATIC CONTROL (2021)

Correction Robotics

LOCUS: A Multi-Sensor Lidar-Centric Solution for High-Precision Odometry and 3D Mapping in Real-Time (vol 6, pg 421, 2020)

Matteo Palieri, Benjamin Morrell, Abhishek Thakur, Kamak Ebadi, Jeremy Nash, Arghya Chatterjee, Christoforos Kanellakis, Luca Carlone, Cataldo Guaragnella, Ali-akbar Agha-mohammadi

IEEE ROBOTICS AND AUTOMATION LETTERS (2021)

Article Robotics

LOCUS: A Multi-Sensor Lidar-Centric Solution for High-Precision Odometry and 3D Mapping in Real-Time

Matteo Palieri, Benjamin Morrell, Abhishek Thakur, Kamak Ebadi, Jeremy Nash, Arghya Chatterjee, Christoforos Kanellakis, Luca Carlone, Cataldo Guaragnella, Ali-akbar Agha-mohammadi

Summary: A high-precision lidar odometry system, named LOCUS, is proposed in this work to achieve robust and real-time operation under challenging perceptual conditions, incorporating a multi-stage scan matching unit and a sensor integration module for seamless fusion.

IEEE ROBOTICS AND AUTOMATION LETTERS (2021)

Article Automation & Control Systems

Robust and Adaptive Sequential Submodular Optimization

Vasileios Tzoumas, Ali Jadbabaie, George J. Pappas

Summary: In this article, the authors propose a robust and adaptive maximization algorithm for solving discrete optimization problems in adversarial environments. The algorithm, called RAM, runs in an online fashion and adapts to the history of failures in each step. It guarantees near-optimal performance and has both provable per-instance a priori bounds and tight and/or optimal a posteriori bounds.

IEEE TRANSACTIONS ON AUTOMATIC CONTROL (2022)

Article Robotics

Resilient Active Information Acquisition With Teams of Robots

Brent Schlotfeldt, Vasileios Tzoumas, George J. Pappas

Summary: This article introduces a robust and adaptive multirobot planning algorithm RAIN, which can plan information acquisition tasks in adversarial environments and exhibits superior performance in multiple information acquisition scenarios.

IEEE TRANSACTIONS ON ROBOTICS (2022)

Article Robotics

Kimera-Multi: Robust, Distributed, Dense Metric-Semantic SLAM for Multi-Robot Systems

Yulun Tian, Yun Chang, Fernando Herrera Arias, Carlos Nieto-Granda, Jonathan P. How, Luca Carlone

Summary: This paper presents $\mathsf {\text{Kimera-Multi}}$, a multi-robot SLAM system that is robust, fully distributed, and capable of capturing semantic information. Experimental results demonstrate its superior performance.

IEEE TRANSACTIONS ON ROBOTICS (2022)

Article Robotics

Distributed Attack-Robust Submodular Maximization for Multirobot Planning

Lifeng Zhou, Vasileios Tzoumas, George J. Pappas, Pratap Tokekar

Summary: In this article, algorithms are designed to protect swarm-robotics applications against sensor denial-of-service attacks, and a distributed robust maximization algorithm is proposed. The distributed approach improves computational speed and achieves tracking performance comparable to centralized algorithms in simulations. Additionally, an improved distributed robust maximization algorithm is introduced to infer attack quantities more accurately.

IEEE TRANSACTIONS ON ROBOTICS (2022)

Article Robotics

LOCUS 2.0: Robust and Computationally Efficient Lidar Odometry for Real-Time 3D Mapping

Andrzej Reinke, Matteo Palieri, Benjamin Morrell, Yun Chang, Kamak Ebadi, Luca Carlone, Ali-Akbar Agha-Mohammadi

Summary: This research presents LOCUS 2.0, a robust and computationally-efficient lidar odometry system for real-time underground 3D mapping. LOCUS 2.0 includes a novel normals-based GICP formulation, an adaptive voxel grid filter, and a sliding-window map approach to reduce computation time and memory consumption. The proposed approach is suitable for use on heterogeneous robotic platforms in large-scale environments with severe computation and memory constraints.

IEEE ROBOTICS AND AUTOMATION LETTERS (2022)

Article Robotics

Loop Closure Prioritization for Efficient and Scalable Multi-Robot SLAM

Christopher E. Denniston, Yun Chang, Andrzej Reinke, Kamak Ebadi, Gaurav S. Sukhatme, Luca Carlone, Benjamin Morrell, Ali-akbar Agha-mohammadi

Summary: This study describes a loop closure module that optimizes the computation of loop closures to maintain a drift-free centralized map. Experimental results demonstrate that the system can generate and maintain a map with low error, and effectively select loop closures.

IEEE ROBOTICS AND AUTOMATION LETTERS (2022)

Article Robotics

Online Submodular Coordination With Bounded Tracking Regret: Theory, Algorithm, and Applications to Multi-Robot Coordination

Zirui Xu, Hongyu Zhou, Vasileios Tzoumas

Summary: We introduce a submodular coordination algorithm with bounded tracking regret to enable multiple robots to coordinate in dynamic, unstructured, and adversarial environments for tasks such as target tracking, environmental mapping, and area monitoring.

IEEE ROBOTICS AND AUTOMATION LETTERS (2023)

Article Engineering, Multidisciplinary

Computation-Communication Trade-Offs and Sensor Selection in Real-Time Estimation for Processing Networks

Luca Ballotta, Luca Schenato, Luca Carlone

IEEE TRANSACTIONS ON NETWORK SCIENCE AND ENGINEERING (2020)

暂无数据