Article
Computer Science, Artificial Intelligence
Tadewos G. Tadewos, Laya Shamgah, Ali Karimoddini
Summary: This paper proposes a systematic approach for automatic tasking and coordination of a heterogeneous team of cooperative autonomous vehicles forming an intelligent vehicle. The approach includes a hierarchical modular coordination algorithm and a hierarchical auctioning algorithm to assign tasks effectively among the vehicles and adapt to operational and duration cost requirements. Different case studies are used to illustrate the developed algorithms.
KNOWLEDGE-BASED SYSTEMS
(2023)
Review
Engineering, Aerospace
Jia Song, Kai Zhao, Yang Liu
Summary: This review article provides an update on the progress of the Mission Planning Problem (MPP) on Multi-UAV, focusing on the burning issue of task assignment. It compares the characteristics of mathematical programming method, heuristic algorithm, negotiation algorithm, and neural networks. The paper discusses different trajectory planning approaches and introduces common collaborative guidance methods. It emphasizes the need for ongoing research, addressing timeliness of task assignment, information coupling, and problems caused by multiple constraints and environmental uncertainty in MPP.
Article
Computer Science, Interdisciplinary Applications
Behrouz Alizadeh Mousavi, Cathal Heavey, Chirine Millauer, Zhikang Tian, Hans Ehm
Summary: This article focuses on designing, developing, and testing a prototype Decision Support System (DSS) based on a mathematical model to assist human planners and improve demand fulfillment systems in tight supply situations, specifically in the semiconductor industry. By incorporating digitalization and supporting human intervention, our approach enhances the complex hierarchical supply chain planning systems. The developed mathematical model, applied as a web application decision support tool called the Regional Customer Allocation Support Tool (ReCAST), is shown to effectively support decision-making processes in a real-world context, as demonstrated through a semiconductor case study.
JOURNAL OF INDUSTRIAL INFORMATION INTEGRATION
(2023)
Article
Transportation Science & Technology
Xi Lin, Chengzhang Wang, Kaiping Wang, Meng Li, Xiangqian Yu
Summary: The study introduces a two-stage control network approach for trajectory planning of UAVs in complex urban environments, addressing the challenges associated with a large number of obstacles. The proposed method efficiently identifies reasonable UAV trajectories within negligible computational time.
TRANSPORTATION RESEARCH PART C-EMERGING TECHNOLOGIES
(2021)
Article
Computer Science, Information Systems
Zhekun Cheng, Liangyu Zhao, Zhongjiao Shi
Summary: In this paper, a decentralized multi-UAV path planning method suitable for obstacle environments is proposed to solve the low computational efficiency and poor scalability of traditional methods. A two-layer coordinative framework is developed to minimize formation rendezvous time and energy consumption. The method is verified through numerical simulations and shows good scalability and potential applications in urban flight and supplies delivery tasks.
Article
Computer Science, Hardware & Architecture
Changhua Yao, Hui Tian, Cong Wang, Liubin Song, Jun Jing, Wenfeng Ma
Summary: This article discusses the challenges and potentials of autonomous UAV swarm control and communication, and introduces a new perspective on joint optimization of these two issues. A task-oriented integration of communication and control (ToICC) framework is proposed, incorporating a mathematization route for task handling capability and a game-based distributed multi-agent cooperative learning approach to enhance the swarm's task handling capability.
IEEE WIRELESS COMMUNICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Chuang Liu, Huaping Chen, Xueping Li, Zeyu Liu
Summary: This paper investigates the optimization problem of using drones for package delivery, reducing the number of drones in the warehouse through a mixed integer programming model and genetic algorithm to ensure timely delivery of packages. The results show that the SDSMGA model outperforms other algorithms as the number of packages increases.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Engineering, Mechanical
Shijia Kang, Peter Xiaoping Liu, Huanqing Wang
Summary: This paper presents a control scheme based on adaptive fuzzy approach for dealing with the finite-time prescribed performance control problem of large-scale nonlinear interconnected systems with input dead zone. By introducing specific functions and design, it solves the issues in conventional backstepping control and ensures that signals in the control system are bounded, with output errors reaching a predefined range within a finite time.
NONLINEAR DYNAMICS
(2021)
Article
Computer Science, Information Systems
Marwan Dhuheir, Emna Baccour, Aiman Erbad, Sinan Sabeeh Al-Obaidi, Mounir Hamdi
Summary: The deployment flexibility and maneuverability of unmanned aerial vehicles (UAVs) have increased their adoption in various applications. However, real-time data processing in UAVs can be challenging due to connection instability and limited bandwidth. To address this issue, this article proposes a model for distributed collaborative inference requests and path planning in a UAV swarm. The model aims to minimize latency by dividing inference requests into multiple parts and planning UAV trajectories to reduce data transmission latency. The proposed model outperforms existing models, as demonstrated through extensive simulations.
IEEE INTERNET OF THINGS JOURNAL
(2023)
Article
Engineering, Electrical & Electronic
Yuchao Zhu, Shaowei Wang
Summary: This paper proposes a cooperative trajectory planning scheme to address the energy issue of unmanned aerial vehicles (UAVs) by using a truck as a mobile recharging station. The scheme aims to minimize the total mission time for gathering data from sensor nodes by dividing the mission area into subregions and finding optimal trajectories for the UAV and the truck. An efficient clustering algorithm is introduced for load-balanced partitioning, and a three-step heuristic algorithm is used to solve the trajectory planning problem. Numerical results demonstrate the effectiveness and cost-efficiency of the proposed scheme for data collection in large-scale wireless sensor networks.
IEEE TRANSACTIONS ON COMMUNICATIONS
(2022)
Article
Automation & Control Systems
Zhihong Liu, Xiangke Wang, Lincheng Shen, Shulong Zhao, Yirui Cong, Jie Li, Dong Yin, Shengde Jia, Xiaojia Xiang
Summary: In this article, a multilayered and distributed architecture for mission-oriented miniature fixed-wing UAV swarms is proposed. The architecture divides the system into five layers and uses interfaces for communication between modules, thereby reducing complexity and supporting diversified missions. Each UAV autonomously performs decision-making procedures, ensuring scalability. Different types of aerial platforms can be easily extended using control allocation matrices and integrated hardware. The proposed architecture is evaluated through field experiments, demonstrating successful formation flight, target recognition, and tracking missions in an integrated architecture for fixed-wing UAV swarms.
IEEE TRANSACTIONS ON SYSTEMS MAN CYBERNETICS-SYSTEMS
(2022)
Article
Agricultural Engineering
Martina Mammarella, Lorenzo Comba, Alessandro Biglia, Fabrizio Dabbene, Paolo Gay
Summary: Fully-autonomous vehicles, both aerial and ground, have great potential in the Agriculture 4.0 framework, especially when operating within cooperative architectures. They are capable of tackling difficult tasks in complex and unstructured scenarios, such as vineyards on sloped terrains. A decentralised multi-phase approach is proposed as an alternative to common cooperative schemes. In this study, the approach is applied to a specific case study involving a vineyard, demonstrating improved autonomous driving capabilities and automated map retrieval for navigation by aerial and ground drones.
BIOSYSTEMS ENGINEERING
(2022)
Article
Automation & Control Systems
Weilin Wang, Yanwei Zang, Shigemasa Takai, Lachlan L. H. Andrew, Chaohui Gong
Summary: This paper addresses the supervisory control of discrete event systems with observation delay and control delay. It extends the existing literature on networked supervisory control by considering varying upper bounds for observation delay and control delay among different event strings, as well as accommodating dynamic observation. The paper proves that satisfying both controllability and delay observability over the combined delay is necessary and sufficient for the existence of a supervisor that can deterministically achieve the desired language. Furthermore, it reduces the problems of networked supervisory control to traditional supervisory control without any delay, and applies the reduction to minimize sensor activations with control delay. The findings of this research have significant implications for the supervisory control of discrete event systems.
Article
Agricultural Engineering
Martina Mammarella, Lorenzo Comba, Alessandro Biglia, Fabrizio Dabbene, Paolo Gay
Summary: Agriculture 4.0 uses technologies such as sensors, information systems, enhanced machinery, and informed management to optimize production by considering variabilities and uncertainties in agricultural systems. This study analyzes and understands the technical and methodological challenges involved, presents cooperative schemes and vehicle models for collaborative machines in agricultural scenarios, and provides an overview of state-of-the-art technologies for autonomous drone guidance. The application of these techniques in a case study in sloped vineyards is also reported.
BIOSYSTEMS ENGINEERING
(2022)
Article
Automation & Control Systems
Mustafa O. Karabag, Melkior Ornik, Ufuk Topcu
Summary: The use of deceptive strategies is important for agents in adversarial environments. This study focuses on the synthesis of optimal deceptive policies and reference policies in a Markov decision process. The synthesis of optimal reference policies is proven to be NP-hard.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL
(2022)
Article
Robotics
Bruno Brito, Michael Everett, Jonathan P. How, Javier Alonso-Mora
Summary: This method learns an interaction-aware policy through deep reinforcement learning to provide long-term guidance for robots, significantly improving robot navigation performance. Compared to previous MPC frameworks, the method performs remarkably well in terms of collision number, travel time, and collisions in cooperative, competitive, and mixed multiagent scenarios.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2021)
Article
Robotics
Yulun Tian, Kasra Khosoussi, David M. Rosen, Jonathan P. How
Summary: This article introduces the first certified correct algorithm for distributed pose-graph optimization, which achieves globally optimal solutions and proposes the distributed Riemannian gradient framework and Riemannian block coordinate descent method. Extensive evaluations show that the proposed method can correctly recover globally optimal solutions under moderate noise.
IEEE TRANSACTIONS ON ROBOTICS
(2021)
Article
Robotics
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
Jesus Tordesillas, Brett T. Lopez, Michael Everett, Jonathan P. How
Summary: This paper introduces FASTER (Fast and Safe Trajectory Planner), which enables UAVs to fly at high speeds in unknown environments while ensuring safety. The algorithm optimizes the local planner in both the free-known and unknown spaces to achieve high-speed trajectories, with a safe back-up trajectory always available in the free-known space.
IEEE TRANSACTIONS ON ROBOTICS
(2022)
Article
Computer Science, Software Engineering
Jesus Tordesillas, Jonathan P. How
Summary: This paper studies the polynomial bases that generate the smallest n-simplex enclosing a given nth-degree polynomial curve in Rn. High-quality feasible solutions are obtained using Sum-Of-Squares programming, branch and bound, and moment relaxations. Global optimality is proven for n = 1, 2, 3.
COMPUTER-AIDED DESIGN
(2022)
Article
Computer Science, Artificial Intelligence
Michael Everett, Bjorn Lutjens, Jonathan P. How
Summary: This work leverages research on certified adversarial robustness to develop a defense mechanism for deep reinforcement learning algorithms that identifies and chooses robust actions in the worst case deviation in input space due to possible adversaries or noise. The approach is demonstrated to increase robustness to noise and adversaries in pedestrian collision avoidance scenarios, a classic control task, and Atari Pong.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2022)
Article
Robotics
Parker C. Lusk, Devarth Parikh, Jonathan P. How
Summary: Using geometric landmarks like lines and planes can improve navigation accuracy and reduce map storage requirements. However, landmark-based registration for loop closure detection is challenging due to the lack of reliable initial guesses. We use the affine Grassmannian manifold to represent 3D lines and planes, proving their rotational and translational invariance, which enables accurate registration using our graph-based data association framework. Our approach outperforms other methods in landmark matching and registration tasks, achieving a 1.7x and 3.5x improvement in successful registrations.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Robotics
Qiangqiang Huang, Can Pu, Kasra Khosoussi, David M. Rosen, Dehann Fourie, Jonathan P. How, John J. Leonard
Summary: This paper introduces NF-iSAM, a novel algorithm for inferring the full posterior distribution in SLAM problems with nonlinear measurement models and non-Gaussian factors. NF-iSAM uses neural networks and normalizing flows to model and sample the full posterior. By leveraging the Bayes tree, NF-iSAM achieves efficient incremental updates in the non-Gaussian setting. Experimental results on range-only SLAM problems with data association ambiguity show that NF-iSAM outperforms state-of-the-art algorithms in accuracy of posterior beliefs for continuous and discrete variables.
IEEE TRANSACTIONS ON ROBOTICS
(2023)
Article
Robotics
Jesus Tordesillas, Jonathan P. How
Summary: This letter presents Deep-PANTHER, a learning-based perception-aware trajectory planner for unmanned aerial vehicles (UAVs) in dynamic environments. It uses imitation learning to train a policy that can generate multiple trajectories to avoid dynamic obstacles while maximizing the field of view of the onboard camera. Extensive simulations demonstrate that Deep-PANTHER outperforms an optimization-based expert in terms of replanning time and loss.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Robotics
Parker C. Lusk, Kaveh Fathian, Jonathan P. How
Summary: We propose a multiway fusion algorithm that can directly handle uncertain pairwise affinities. Instead of relying on initial pairwise associations like existing methods, our MIXER algorithm improves accuracy by utilizing the additional information provided by pairwise affinities. Our main contribution is a multiway fusion formulation suitable for non-binary affinities and a novel continuous relaxation that guarantees binary solutions, avoiding problematic solution binarization steps.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Robotics
Xiaoyi Cai, Brent Schlotfeldt, Kasra Khosoussi, Nikolay Atanasov, George J. Pappas, Jonathan P. How
Summary: This article addresses the problem of coordinating sensor-equipped robots in a safe manner to reduce uncertainty in a dynamical process, considering the tradeoff between information gain and energy cost. Existing multirobot planners based on coordinate descent fail to provide performance guarantees due to the nonmonotone objective function in robot trajectories. Methods that handle nonmonotonicity also lose their performance guarantees when collision avoidance constraints are imposed. To achieve both performance and safety guarantees, this work proposes a hierarchical approach that combines a distributed planner with worst-case performance guarantees and a decentralized controller based on control barrier functions. Extensive simulations, hardware-in-the-loop tests, and hardware experiments demonstrate that the proposed approach achieves a better tradeoff between sensing and energy cost compared to coordinate-descent-based algorithms.
IEEE TRANSACTIONS ON ROBOTICS
(2023)
Article
Computer Science, Information Systems
Jesus Tordesillas, Jonathan P. How
Summary: This paper presents PANTHER, a real-time perception-aware trajectory planner for multirotor-UAVs in dynamic environments. PANTHER plans trajectories that avoid dynamic obstacles while keeping them in the sensor field of view and minimizing blur for object tracking. The joint optimization of UAV rotation and translation allows PANTHER to fully exploit the differential flatness of multirotors. The experiments show that PANTHER outperforms existing methods in multi-obstacle avoidance scenarios.
Article
Computer Science, Information Systems
Sharan Raja, Golnaz Habibi, Jonathan P. How
Summary: This paper introduces a deep reinforcement learning framework, CA-CBBA, for task allocation in environments with limited bandwidth and message collisions. By learning communication strategies, this approach significantly improves task allocation performance and outperforms baseline algorithms in various scenarios.
Article
Automation & Control Systems
Michael Everett, Golnaz Habibi, Jonathan P. How
Summary: Neural networks are commonly used in uncertain environments, but there is a lack of tools for formally analyzing the uncertainty in their outputs. Recent works have attempted to approximate the propagation of uncertainty through nonlinear activations, but this has led to excessive conservatism and slow computation speeds. This letter proposes a unified approach that combines propagation and partition methods to provide tighter bounds on neural network outputs.
IEEE CONTROL SYSTEMS LETTERS
(2021)
Article
Computer Science, Information Systems
Michael Everett, Yu Fan Chen, Jonathan P. How
Summary: This work introduces a deep reinforcement learning algorithm to enable robots to avoid collisions with pedestrians and other agents, even in heterogeneous and non-communicative environments. By leveraging Long Short-Term Memory strategy, the algorithm can adapt to varying scenarios and observe a larger number of agents, rather than just a fixed number of neighbors.