Logo V Conferencia Colombiana  de Control Automático - CCAC, CCAC 2021, IEEE CCAC

5th IEEE Colombian Conference on Automatic Control (CCAC)

October
19th - 22nd / 2021

Technological advances for sustainable regional development

Workshops

Workshop # 1:

Fast actuation and adaptation in Aerial Modular Robots

Is an Assistant Professor in Computer Science and Engineering at Lehigh University. He worked as a Post-Doctoral Researcher at the GRASP Laboratory, University of Pennsylvania. His main research is focused on modular aerial robots, multi-robot systems, and robot swarms. He received his B.Sc. (2010) and M.Sc. (2012) in Informatics Engineering from National University of Colombia, and his Ph.D. in Computer Science, Artificial Intelligence and Robotics at UFMG, Brazil (2017). His current projects include enhancing resilience in large-scale robot networks, and co-designing algorithms, dynamical models, and robot hardware for modular robots.

Abstract:
actuacion y adaptacionDuring emergencies in urban scenarios, tasks like manipulation and transportation need to be performed rapidly to save human lives. Instead of using large and task-specific robots, we propose modular robot swarms, composed of hundreds of aerial modules that can rapidly adapt to achieve aerial tasks. Those modular robots must be able to rapidly change their shape and actuation capabilities to perform adaptable operations in time-critical situations. The potential and flexibility of the modular aerial systems are mainly in their ability to change their shape by creating and removing physical connections between modules. In this workshop, we describe the main concepts in this area and our recent results on aerial vehicles that can self-assemble, self-adapt, and self-heal in midair. Our work is mainly focused on co-designing algorithms, dynamical models, and robot hardware that enhances adaptability, scalability, and resiliency.
David Saldaña
David Saldaña/ Lehigh University /

Workshop # 2:

Optimal structured control in spatially-distributed systems: convexity, performance and decentralization.

Is a PhD candidate in Applied Mathematics at MIT, with a previous background in engineering. Her research revolves around the design of optimal structured controllers and estimators for spatially-distributed systems, with particular emphasis on systems in which the dynamics are described by PDEs. Often, she draws inspiration from biology to come up with novel problem formulations and control architectures. She has received several awards and recognitions for her academic work, including the National Award for Academic Excellence from the Government of Spain, a Google scholarship, and a Presidential Fellowship from MIT.

Abstract:

estructura optima de control

Large-scale networks of spatially distributed dynamical systems are common in modern applications. For these, traditional optimal control design techniques, such as Linear Quadratic Regulation (LQR), usually yield all-to-all (centralized) communication topologies which are often prohibitive due to their large scale. This challenge justifies the increasing interest of the research community in the design of optimal spatially structured controllers, with limited information exchange between subsystems.

In the first part of this workshop, we will briefly cover the well-known LQR problem and study properties of its solution. Afterwards, we will present some approaches to the design of optimal spatially structured controllers and discuss associated challenges. In the second part of the workshop, we will focus on systems in which the underlying dynamics are spatially invariant (SI), a particular class of spatially distributed systems, suitable to model large arrays of identical subsystems. We will study the spatial localization properties of optimal controllers for SI systems and introduce a convex formulation for the design of optimal controllers with spatial structural constraints. We will provide some examples to ease understanding.

 

Juncal Arbelaiz
Juncal Arbelaiz / Massachusetts Institute of technology – MIT /

Workshop # 3:

Decentralized algorithms for collaborative learning

Is an assistant professor in the Department of Electrical Engineering at The Pennsylvania State University. She received the B.E. degree in electronic information from the Huazhong University of Science and Technology, Wuhan, China, in 2011, and the Ph.D. degree in electronic and computer engineering from The Hong Kong University of Science and Technology in 2016. She was a postdoc researcher with the School of Industrial Engineering, Purdue University from 2016 to 2020. Her research interests include statistical signal processing, optimization algorithms and machine learning. She is a co-recipient of a student best paper at IEEE International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP) 2017, and a recipient of the 2020 IEEE Signal Processing Society Young Author Best Paper Award.

Abstract:

computacion

Advances in computation, communication, and data storage techniques in recent decades significantly reduced the cost of data acquisition, leading to an explosion of data generated across different interconnected platforms. Apart from the computational difficulties that arise from non-convex formulations, the sheer volume and spatial disparity of data also pose challenges to traditional learning procedures, which typically require centralized training sets. Reaping the dividend offered by the data deluge necessitates the development of collaborative learning methods capable of making inferences from data over the network.

This workshop discusses some recent developments in decentralized learning algorithms and their computational and statistical guarantees. A quick introduction to distributed optimization will be given in the first part of the workshop. Then we will present a novel algorithmic framework, SONATA, and its convergence for learning problems in different classes. The second part of the workshop focuses on the statistical properties of the algorithms.  While statistical-computation tradeoffs have been largely explored in the centralized setting, our understanding over meshed networks is limited. Distributed schemes, designed and performing well in the classical low-dimensional regime, can break down in the high-dimensional case; and existing convergence studies may fail to predict algorithmic behaviors. This is mainly because most distributed algorithms have been designed and studied only from the optimization perspective, lacking the statistical dimension. Through some vignettes from low- and high-dimensional statistical inference, we will go over some designs and new analyses aiming at bringing statistical thinking in distributed optimization.

 

Ying Sun
Ying Sun/ Pennsylvania State University /

Workshop # 4:

Robust optimization in cyber-physical systems with applications in electricity demand response

Mahnoosh Alizadeh
Mahnoosh Alizadeh/ UC Santa Bárbara /

Is an assistant professor of Electrical and Computer Engineering at the University of California Santa Barbara. She received the B.Sc. degree (’09) in Electrical Engineering from Sharif University of Technology and the M.Sc. (’13) and Ph.D. (’14) degrees in Electrical and Computer Engineering from the University of California Davis. From 2014 to 2016, she was a postdoctoral scholar at Stanford University. Her research interests are focused on designing network control, optimization, and learning frameworks to promote efficiency and resiliency in societal-scale cyber-physical systems. Dr. Alizadeh is a recipient of the NSF CAREER award.

Berkay Turan
Berkay Turan / UC Santa Bárbara /

Is pursuing the Ph.D. degree in Electrical and Computer Engineering at the University of California, Santa Barbara. He received the B.Sc. degree in Electrical and Electronics Engineering and the B.Sc. degree in Physics from Bogazici University, Istanbul, Turkey, in 2018. His research interests focus on designing efficient control mechanisms and robust distributed optimization methods for cyber-physical systems.  

Abstract:

secuencia optimizacion

Distributed optimization algorithms have found wide applications in many cyber-physical systems where it is not possible to centrally coordinate the behavior of multiple agents towards optimal system operation. Examples include power system optimization, network routing, or robot formation control. However, classical distributed optimization algorithms require all communications between the agents and a central coordinator to be trustworthy, making them susceptible to failure due to many factors such as adversarial influences, component failures, or miscommunication. Consider for example the case of residential demand management in power system optimization through home energy management (HEM) systems. Should a few HEM units in a neighborhood become compromised by an adversary, distributed energy management algorithms may fail to converge, or even worse, power system constraints might become violated, potentially leading to grid failures. Motivated by this challenge, this workshop will 1) briefly introduce a number of representative distributed optimization algorithms in cyber-physical systems; 2) discuss the importance of designing distributed optimization methods that are robust to arbitrary and potentially adversarial corruption of the agents for distributed resource allocation and distributed learning problems; 3) introduce a new robust primal-dual algorithm for distributed resource allocation problems. Because resource allocation schemes operate on actual physical infrastructure, a challenging task for robustness is to guarantee the safety of the system despite adversarial influences. 4) Lastly, we will demonstrate that robust temporal averaging significantly improves performance in many distributed resource allocation and distributed learning problems subject to adversarial corruption.

 

Workshop # 5:

Resilient distributed machine learning: Secure multi-agent federation

An Assistant Professor in the Dept. of Electrical & Computer Engineering with a courtesy appointment in Khoury College of Computer Sciences at Northeastern University. Prior joining Northeastern, she was a Postdoc in MIT Computer Science & Artificial Intelligence Laboratory (CSAIL) from 2017-2020, hosted by Prof. Nancy Lynch. She obtained her Ph.D and M.S. degree from Dept. of Electrical & Computer Engineering at University of Illinois at Urbana-Champaign in 2017 and 2014, respectively. Her research intersects machine learning, security & privacy, and distributed systems. She was a finalist (1/3) for the Best Student Paper Award at DISC 2016, and she received the 2015 Best Student Paper Award at SSS 2015. She received UIUC's Sundaram Seshu International Student Fellowship for 2016 and was invited to participate in Rising Stars in EECS(2018). She has served on TPC for several flagship conferences including DISC, SIGMETRICS and ICDCS.

Abstract:

aprendizaje automatico

With the rapid increasing data-collection, storage, and computation capabilities of personal computing devices such as laptops and smartphones, and with the growing popularity of wearable devices such as Apple Watch, one recent trend in machine learning is to outsource part of the involved computation burden to external edge and/or end devices; in a sense, the edge and end devices can be viewed as external workers of the cloud. 

In this workshop, we will introduce Federated Learning - a practical learning paradigm wherein the training data is kept confidentially on users' own end devices. Then we will talk about the vulnerabilities of FL to different types of internal and external failures and attacks -- formal and rigorous mathematical models will be provided. We will talk about several state-of-the-art approaches to tackle this variety of vulnerabilities, starting from the relatively simple distributed context awareness problems to the more advanced Byzantine-resilient regression problems. Specifically, we will talk about how to be able to handle the local data sparsity, measurement contamination, arbitrarily malicious misleading messages, adversary collusion, etc. 

 

Lili Sue
Lili Sue/ Northeastern University /
© 2022 Conferencia Colombiana de Control Automático - CCAC. Todos los derechos reservados.