arXiv - CS - Systems and Control最新文献_第4页

Obstacle Avoidance of Autonomous Vehicles: An LPVMPC with Scheduling Trust Region 自动驾驶车辆的避障：具有调度信任区域的 LPVMPC

arXiv - CS - Systems and Control Pub Date : 2024-05-03 DOI: arxiv-2405.02030

Maryam Nezami, Dimitrios S. Karachalios, Georg Schildbach, Hossam S. Abbas

{"title":"Obstacle Avoidance of Autonomous Vehicles: An LPVMPC with Scheduling Trust Region","authors":"Maryam Nezami, Dimitrios S. Karachalios, Georg Schildbach, Hossam S. Abbas","doi":"arxiv-2405.02030","DOIUrl":"https://doi.org/arxiv-2405.02030","url":null,"abstract":"Reference tracking and obstacle avoidance rank among the foremost challenging\u0000aspects of autonomous driving. This paper proposes control designs for solving\u0000reference tracking problems in autonomous driving tasks while considering\u0000static obstacles. We suggest a model predictive control (MPC) strategy that\u0000evades the computational burden of nonlinear nonconvex optimization methods\u0000after embedding the nonlinear model equivalently to a linear parameter-varying\u0000(LPV) formulation using the so-called scheduling parameter. This allows optimal\u0000and fast solutions of the underlying convex optimization scheme as a quadratic\u0000program (QP) at the expense of losing some performance due to the uncertainty\u0000of the future scheduling trajectory over the MPC horizon. Also, to ensure that\u0000the modeling error due to the application of the scheduling parameter\u0000predictions does not become significant, we propose the concept of scheduling\u0000trust region by enforcing further soft constraints on the states and inputs. A\u0000consequence of using the new constraints in the MPC is that we construct a\u0000region in which the scheduling parameter updates in two consecutive time\u0000instants are trusted for computing the system matrices, and therefore, the\u0000feasibility of the MPC optimization problem is retained. We test the method in\u0000different scenarios and compare the results to standard LPVMPC as well as\u0000nonlinear MPC (NMPC) schemes.","PeriodicalId":501062,"journal":{"name":"arXiv - CS - Systems and Control","volume":"37 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-05-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140882318","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Physics-informed generative neural networks for RF propagation prediction with application to indoor body perception 用于射频传播预测的物理信息生成神经网络在室内人体感知中的应用

arXiv - CS - Systems and Control Pub Date : 2024-05-03 DOI: arxiv-2405.02131

Federica Fieramosca, Vittorio Rampa, Michele D'Amico, Stefano Savazzi

引用次数: 0

Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots 学习轮足机器人的鲁棒自主导航和运动能力

arXiv - CS - Systems and Control Pub Date : 2024-05-03 DOI: arxiv-2405.01792

Joonho Lee, Marko Bjelonic, Alexander Reske, Lorenz Wellhausen, Takahiro Miki, Marco Hutter

{"title":"Learning Robust Autonomous Navigation and Locomotion for Wheeled-Legged Robots","authors":"Joonho Lee, Marko Bjelonic, Alexander Reske, Lorenz Wellhausen, Takahiro Miki, Marco Hutter","doi":"arxiv-2405.01792","DOIUrl":"https://doi.org/arxiv-2405.01792","url":null,"abstract":"Autonomous wheeled-legged robots have the potential to transform logistics\u0000systems, improving operational efficiency and adaptability in urban\u0000environments. Navigating urban environments, however, poses unique challenges\u0000for robots, necessitating innovative solutions for locomotion and navigation.\u0000These challenges include the need for adaptive locomotion across varied\u0000terrains and the ability to navigate efficiently around complex dynamic\u0000obstacles. This work introduces a fully integrated system comprising adaptive\u0000locomotion control, mobility-aware local navigation planning, and large-scale\u0000path planning within the city. Using model-free reinforcement learning (RL)\u0000techniques and privileged learning, we develop a versatile locomotion\u0000controller. This controller achieves efficient and robust locomotion over\u0000various rough terrains, facilitated by smooth transitions between walking and\u0000driving modes. It is tightly integrated with a learned navigation controller\u0000through a hierarchical RL framework, enabling effective navigation through\u0000challenging terrain and various obstacles at high speed. Our controllers are\u0000integrated into a large-scale urban navigation system and validated by\u0000autonomous, kilometer-scale navigation missions conducted in Zurich,\u0000Switzerland, and Seville, Spain. These missions demonstrate the system's\u0000robustness and adaptability, underscoring the importance of integrated control\u0000systems in achieving seamless navigation in complex environments. Our findings\u0000support the feasibility of wheeled-legged robots and hierarchical RL for\u0000autonomous navigation, with implications for last-mile delivery and beyond.","PeriodicalId":501062,"journal":{"name":"arXiv - CS - Systems and Control","volume":"18 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-05-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140882685","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Reinforcement Learning control strategies for Electric Vehicles and Renewable energy sources Virtual Power Plants 电动汽车和可再生能源虚拟发电厂的强化学习控制策略

arXiv - CS - Systems and Control Pub Date : 2024-05-03 DOI: arxiv-2405.01889

Francesco Maldonato, Izgh Hadachi

{"title":"Reinforcement Learning control strategies for Electric Vehicles and Renewable energy sources Virtual Power Plants","authors":"Francesco Maldonato, Izgh Hadachi","doi":"arxiv-2405.01889","DOIUrl":"https://doi.org/arxiv-2405.01889","url":null,"abstract":"The increasing demand for direct electric energy in the grid is also tied to\u0000the increase of Electric Vehicle (EV) usage in the cities, which eventually\u0000will totally substitute combustion engine Vehicles. Nevertheless, this high\u0000amount of energy required, which is stored in the EV batteries, is not always\u0000used and it can constitute a virtual power plant on its own. Bidirectional EVs\u0000equipped with batteries connected to the grid can therefore charge or discharge\u0000energy depending on public needs, producing a smart shift of energy where and\u0000when needed. EVs employed as mobile storage devices can add resilience and\u0000supply/demand balance benefits to specific loads, in many cases as part of a\u0000Microgrid (MG). Depending on the direction of the energy transfer, EVs can\u0000provide backup power to households through vehicle-to-house (V2H) charging, or\u0000storing unused renewable power through renewable-to-vehicle (RE2V) charging.\u0000V2H and RE2V solutions can complement renewable power sources like solar\u0000photovoltaic (PV) panels and wind turbines (WT), which fluctuate over time,\u0000increasing the self-consumption and autarky. The concept of distributed energy\u0000resources (DERs) is becoming more and more present and requires new solutions\u0000for the integration of multiple complementary resources with variable supply\u0000over time. The development of these ideas is coupled with the growth of new AI\u0000techniques that will potentially be the managing core of such systems. Machine\u0000learning techniques can model the energy grid environment in such a flexible\u0000way that constant optimization is possible. This fascinating working principle\u0000introduces the wider concept of an interconnected, shared, decentralized grid\u0000of energy. This research on Reinforcement Learning control strategies for\u0000Electric Vehicles and Renewable energy sources Virtual Power Plants focuses on\u0000providing solutions for such energy supply optimization models.","PeriodicalId":501062,"journal":{"name":"arXiv - CS - Systems and Control","volume":"30 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-05-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140882320","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

CGD: Constraint-Guided Diffusion Policies for UAV Trajectory Planning CGD：用于无人飞行器轨迹规划的约束引导扩散策略

arXiv - CS - Systems and Control Pub Date : 2024-05-02 DOI: arxiv-2405.01758

Kota Kondo, Andrea Tagliabue, Xiaoyi Cai, Claudius Tewari, Olivia Garcia, Marcos Espitia-Alvarez, Jonathan P. How

{"title":"CGD: Constraint-Guided Diffusion Policies for UAV Trajectory Planning","authors":"Kota Kondo, Andrea Tagliabue, Xiaoyi Cai, Claudius Tewari, Olivia Garcia, Marcos Espitia-Alvarez, Jonathan P. How","doi":"arxiv-2405.01758","DOIUrl":"https://doi.org/arxiv-2405.01758","url":null,"abstract":"Traditional optimization-based planners, while effective, suffer from high\u0000computational costs, resulting in slow trajectory generation. A successful\u0000strategy to reduce computation time involves using Imitation Learning (IL) to\u0000develop fast neural network (NN) policies from those planners, which are\u0000treated as expert demonstrators. Although the resulting NN policies are\u0000effective at quickly generating trajectories similar to those from the expert,\u0000(1) their output does not explicitly account for dynamic feasibility, and (2)\u0000the policies do not accommodate changes in the constraints different from those\u0000used during training. To overcome these limitations, we propose Constraint-Guided Diffusion (CGD),\u0000a novel IL-based approach to trajectory planning. CGD leverages a hybrid\u0000learning/online optimization scheme that combines diffusion policies with a\u0000surrogate efficient optimization problem, enabling the generation of\u0000collision-free, dynamically feasible trajectories. The key ideas of CGD include\u0000dividing the original challenging optimization problem solved by the expert\u0000into two more manageable sub-problems: (a) efficiently finding collision-free\u0000paths, and (b) determining a dynamically-feasible time-parametrization for\u0000those paths to obtain a trajectory. Compared to conventional neural network\u0000architectures, we demonstrate through numerical evaluations significant\u0000improvements in performance and dynamic feasibility under scenarios with new\u0000constraints never encountered during training.","PeriodicalId":501062,"journal":{"name":"arXiv - CS - Systems and Control","volume":"35 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-05-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140882330","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Addressing the Load Estimation Problem: Cell Switching in HAPS-Assisted Sustainable 6G Networks 解决负载估计问题：HAPS 辅助可持续 6G 网络中的小区切换

arXiv - CS - Systems and Control Pub Date : 2024-05-02 DOI: arxiv-2405.01690

Maryam Salamatmoghadasi, Metin Ozturk, Halim Yanikomeroglu

{"title":"Addressing the Load Estimation Problem: Cell Switching in HAPS-Assisted Sustainable 6G Networks","authors":"Maryam Salamatmoghadasi, Metin Ozturk, Halim Yanikomeroglu","doi":"arxiv-2405.01690","DOIUrl":"https://doi.org/arxiv-2405.01690","url":null,"abstract":"This study aims to introduce and address the problem of traffic load\u0000estimation in the cell switching concept within the evolving landscape of\u0000vertical heterogeneous networks (vHetNets). The problem is that the practice of\u0000cell switching faces a significant challenge due to the lack of accurate data\u0000on the traffic load of sleeping small base stations (SBSs). This problem makes\u0000the majority of the studies in the literature, particularly those employing\u0000load-dependent approaches, impractical due to their basic assumption of perfect\u0000knowledge of the traffic loads of sleeping SBSs for the next time slot. Rather\u0000than developing another advanced cell switching algorithm, this study\u0000investigates the impacts of estimation errors and explores possible solutions\u0000through established methodologies in a novel vHetNet environment that includes\u0000the integration of a high altitude platform (HAPS) as a super macro base\u0000station (SMBS) into the terrestrial network. In other words, this study adopts\u0000a more foundational perspective, focusing on eliminating a significant obstacle\u0000for the application of advanced cell switching algorithms. To this end, we\u0000explore the potential of three distinct spatial interpolation-based estimation\u0000schemes: random neighboring selection, distance-based selection, and\u0000clustering-based selection. Utilizing a real dataset for empirical validations,\u0000we evaluate the efficacy of our proposed traffic load estimation schemes. Our\u0000results demonstrate that the multi-level clustering (MLC) algorithm performs\u0000exceptionally well, with an insignificant difference (i.e., 0.8%) observed\u0000between its estimated and actual network power consumption, highlighting its\u0000potential to significantly improve energy efficiency in vHetNets.","PeriodicalId":501062,"journal":{"name":"arXiv - CS - Systems and Control","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-05-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140882317","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Accounting for the Effects of Probabilistic Uncertainty During Fast Charging of Lithium-ion Batteries 考虑锂离子电池快速充电过程中的概率不确定性影响

arXiv - CS - Systems and Control Pub Date : 2024-05-02 DOI: arxiv-2405.01681

Minsu Kim, Joachim Schaeffer, Marc D. Berliner, Berta Pedret Sagnier, Rolf Findeisen, Richard D. Braatz

{"title":"Accounting for the Effects of Probabilistic Uncertainty During Fast Charging of Lithium-ion Batteries","authors":"Minsu Kim, Joachim Schaeffer, Marc D. Berliner, Berta Pedret Sagnier, Rolf Findeisen, Richard D. Braatz","doi":"arxiv-2405.01681","DOIUrl":"https://doi.org/arxiv-2405.01681","url":null,"abstract":"Batteries are nonlinear dynamical systems that can be modeled by Porous\u0000Electrode Theory models. The aim of optimal fast charging is to reduce the\u0000charging time while keeping battery degradation low. Most past studies assume\u0000that model parameters and ambient temperature are a fixed known value and that\u0000all PET model parameters are perfectly known. In real battery operation,\u0000however, the ambient temperature and the model parameters are uncertain. To\u0000ensure that operational constraints are satisfied at all times in the context\u0000of model-based optimal control, uncertainty quantification is required. Here,\u0000we analyze optimal fast charging for modest uncertainty in the ambient\u0000temperature and 23 model parameters. Uncertainty quantification of the battery\u0000model is carried out using non-intrusive polynomial chaos expansion and the\u0000results are verified with Monte Carlo simulations. The method is investigated\u0000for a constant current--constant voltage charging strategy for a battery for\u0000which the strategy is known to be standard for fast charging subject to\u0000operating below maximum current and charging constraints. Our results\u0000demonstrate that uncertainty in ambient temperature results in violations of\u0000constraints on the voltage and temperature. Our results identify a subset of\u0000key parameters that contribute to fast charging among the overall uncertain\u0000parameters. Additionally, it is shown that the constraints represented by\u0000voltage, temperature, and lithium-plating overpotential are violated due to\u0000uncertainties in the ambient temperature and parameters. The C-rate and charge\u0000constraints are then adjusted so that the probability of violating the\u0000degradation acceleration condition is below a pre-specified value. This\u0000approach demonstrates a computationally efficient approach for determining\u0000fast-charging protocols that take probabilistic uncertainties into account.","PeriodicalId":501062,"journal":{"name":"arXiv - CS - Systems and Control","volume":"61 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-05-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140882201","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Temporal assessment of malicious behaviors: application to turnout field data monitoring 恶意行为的时间评估：应用于道岔现场数据监测

arXiv - CS - Systems and Control Pub Date : 2024-05-02 DOI: arxiv-2405.02346

Sara Abdellaoui, Emil Dumitrescu, Cédric Escudero, Eric Zamaï

引用次数: 0

A Cloud-Edge Framework for Energy-Efficient Event-Driven Control: An Integration of Online Supervised Learning, Spiking Neural Networks and Local Plasticity Rules 高能效事件驱动控制的云边缘框架：在线监督学习、尖峰神经网络和局部可塑性规则的整合

arXiv - CS - Systems and Control Pub Date : 2024-04-12 DOI: arxiv-2405.02316

Reza Ahmadvand, Sarah Safura Sharif, Yaser Mike Banad

{"title":"A Cloud-Edge Framework for Energy-Efficient Event-Driven Control: An Integration of Online Supervised Learning, Spiking Neural Networks and Local Plasticity Rules","authors":"Reza Ahmadvand, Sarah Safura Sharif, Yaser Mike Banad","doi":"arxiv-2405.02316","DOIUrl":"https://doi.org/arxiv-2405.02316","url":null,"abstract":"This paper presents a novel cloud-edge framework for addressing computational\u0000and energy constraints in complex control systems. Our approach centers around\u0000a learning-based controller using Spiking Neural Networks (SNN) on physical\u0000plants. By integrating a biologically plausible learning method with local\u0000plasticity rules, we harness the efficiency, scalability, and low latency of\u0000SNNs. This design replicates control signals from a cloud-based controller\u0000directly on the plant, reducing the need for constant plant-cloud\u0000communication. The plant updates weights only when errors surpass predefined\u0000thresholds, ensuring efficiency and robustness in various conditions. Applied\u0000to linear workbench systems and satellite rendezvous scenarios, including\u0000obstacle avoidance, our architecture dramatically lowers normalized tracking\u0000error by 96% with increased network size. The event-driven nature of SNNs\u0000minimizes energy consumption, utilizing only about 111 nJ (0.3% of conventional\u0000computing requirements). The results demonstrate the system's adjustment to\u0000changing work environments and its efficient use of computational and energy\u0000resources, with a moderate increase in energy consumption of 27.2% and 37% for\u0000static and dynamic obstacles, respectively, compared to non-obstacle scenarios.","PeriodicalId":501062,"journal":{"name":"arXiv - CS - Systems and Control","volume":"61 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2024-04-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"140882510","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Exploring Lightweight Federated Learning for Distributed Load Forecasting 探索用于分布式负荷预测的轻量级联合学习

arXiv - CS - Systems and Control Pub Date : 2024-04-04 DOI: arxiv-2404.03320

Abhishek Duttagupta, Jin Zhao, Shanker Shreejith

引用次数: 0