1. Frank Allgöwer, University of Stuttgart, An Internal Model Principle For Consensus in Heterogeneous Multi-agent Systems.

  2. John Baillieul, Boston University, Brockett-Heisenberg Systems and Geometric Aspects of Control Communication Complexity.

  3. Alexandre M. Bayen, University of California at Berkeley, Real-time Estimation of Distributed Parameters Systems: Application to Large Scale Infrastructure Systems.

  4. Jorge Cortés, University of California at San Diego, Self-Triggered Coordination of Robotic Networks for Optimal Deployment.

  5. Bruce A. Francis, University of Toronto, Infinite Chains of Vehicles.

  6. Laura Giarré, University of Palermo, Resource Sharing in WiFi Infrastructure Networks and Infrastructure-less Ad-Hoc Networks.

  7. Christoforos N. Hadjicostis, University of Cyprus, Distributed Weight Balancing in Directed Graphs.

  8. J. Karl Hedrick, University of California at Berkeley, Prey Modeling in Predatory/Prey Interaction – Risk avoidance, Group Foraging, and Communication.

  9. Steve H. Low, California Institute of Technology, Multi-period Optimal Procurement and Demand Response in the Presence of Uncertain Supply.

  10. Sonia Martínez, University of California at San Diego, Distributed Constrained Optimization under Time-Varying Multi-agent Interactions.

  11. Mehran Mesbahi, University of Washington, Viewing Networks as Systems.

  12. Sean Meyn, University of Illinois at Urbana-Champaign, Mean-field Games for Fun and Profit.

  13. Amit K. Roy-Chowdhury, University of California at Riverside, Integrated Sensing and Analysis in Distributed Camera Networks.

  14. Bruno Sinopoli, Carnegie Mellon University, Asymptotic Performance of Distributed Detection over Random Networks.

  15. Gaurav S. Sukhatme, University of Southern California, Planning and Decision-Making for Underwater Robot Teams: Algorithms and Experiments.


Frank Allgöwer An Internal Model Principle for Consensus in Heterogeneous Multi-agent Systems.

In this presentation we will investigate the problem of synchronizing group of heterogeneous nonlinear dynamical systems that are connected by means of generalized diffusive couplings. The focus is on the derivation of necessary conditions for the existence of such a solution despite the systems possessing non-identical models. We show that for the problem to be solvable a synchronous steady state needs to exist. This condition leads to the requirement that all individual system models need to embed an internal model of some virtual exosystem. The latter condition is expressed in terms of nonlinear partial differential equations that are related to those known from the theory of output regulation.

John Baillieul Brockett-Heisenberg Systems and Geometric Aspects of Control Communication Complexity.

The interaction of information and control has been a topic of interest to system theorists that can be traced back to the 1950’s when the fields of communications, control, and information theory were new but developing rapidly. Recent advances in our understanding of this interplay have emerged from work on the dynamical effect of state quantization and a corresponding understanding of how communication channel data rates affect system stability. While a large body of research has now emerged dealing with communication constrained feedback channels and optimal design of information flows in networks, less attention has been paid to ways in which control systems should be designed in order to optimally mediate computation and communication. Recently W.S. Wong has proposed the concept of control communication complexity (CCC) as a formal approach for understanding how a group of distributed agents can take independent actions that cooperatively realize common goals and objectives. A prototypical goal is the computation of a function, and CCC provides a promising new approach to understanding complexity in terms of the cost of information processing. This lecture will introduce control communication complexity in terms of what are called standard parts optimal control problems. Such optimization problems are of interest in the context of quantum computing, and similar problems have recently been discussed in connection with protocols for assembly of molecular components in synthetic biology.

Alexandre M. Bayen Real-time Estimation of Distributed Parameters Systems: Application to Large Scale Infrastructure Systems.

The coupling of the physical world with information technology promises to help meet increasing demands for efficient, sustainable, and secure management of our built infrastructure and natural environment. A mathematical abstraction of the physical environment can be achieved in the form of distributed parameters systems, described by partial differential equations. Yet, initial and boundary conditions, and other model parameters necessary for complete characterization of these models are often unknown, driving the need for distributed sensing of the physical environment. Because of the nonlinearities and distributed nature inherent to these physical processes, efficient estimation algorithms to reconcile modeling and measurement errors in real-time remains an open challenge for many applications.

This work investigates the problem of real-time estimation of distributed parameters systems in the context of monitoring traffic, river flows and earthquakes. The recent explosion of smartphones with Internet connectivity, GPS and magnetometers is rapidly increasing sensing capabilities for numerous infrastructure systems. The talk will present theoretical results, algorithms and implementations designed to integrate mobile measurements obtained from smartphones into distributed parameter models of infrastructure systems. The models considered include Hamilton-Jacobi equations, first order conservation laws and systems of conservation laws. A new convex formulation of data assimilation and data reconciliation problems will be derived and demonstrated for some of these models. Other techniques developed will be briefly presented as well, relying on ensemble Kalman filtering.

The talk will focus mainly on a traffic monitoring system launched jointly by UC Berkeley and Nokia, called Mobile Millennium, which is operational in Northern California and streams more than 60 million data points a day into traffic models. The talk will also present two more recent applications of this research: the floating sensor network, for real-time riverflow reconstruction, and the iShake system, for smartphone-based real-time earthquake monitoring.

Jorge Cortés Self-Triggered Coordination of Robotic Networks for Optimal Deployment.

This talk studies a deployment problem for a group of robots where individual agents operate with outdated information about each other, locations. The objective is to understand to what extent outdated information is still useful and at which point it becomes essential to obtain new, up-to-date information. We propose a self-triggered coordination algorithm based on spatial partitioning techniques with uncertain information. We analyze its correctness in synchronous and asynchronous scenarios, and establish the same convergence guarantees that a synchronous algorithm with perfect information at all times would achieve. The technical approach combines computational geometry, set-valued stability analysis, and event-based systems.

Bruce Francis Infinite Chains of Vehicles.

In studying the formation of a very large number of vehicles, one approach is instead to model an infinite number of vehicles. This line of research goes back at least to Levine and Athans in 1967, includes Bamieh, Paganini and Dahleh in 2002, then D’Andrea and Dullerud in 2003, and up to Curtain, Iftime and Zwart in 2010. The relevant question is what mathematical framework to take so that the infinite-chain model correctly describes the behaviour of the large-but-finite chain model. The works just mentioned take the state space to be the Hilbert space of square-summable sequences. The advantage is that there is a rich Fourier theory available if the formation is spatially invariant. But this Hilbert space formulation leads to anomalous behaviour. For example, an infinite chain of vehicles when displaced will return to their starting points even though the vehicles do not have global sensing capability and therefore could not in reality do so. This talk proposes a different mathematical framework and describes the progress made so far. This is joint work with Avraham Feintuch, Math Department, Ben Gurion University of the Negev.

Laura Giarré Resource Sharing in WiFi Infrastructure Networks and Infrastructure-less ad-hoc networks.

We present some recent results on resource sharing in WiFi infrastructure networks and infrastructure-less ad-hoc networks.

A game theoretical approach is first considered to analyze and propose some protocol extensions inducing a fair resource sharing for infrastructure networks. Infrastructure networks are characterized by a star topology, which connects multiple mobile nodes to a common station called Access Point (AP). The key idea is using the central role for the access point for implementing a mechanism design scheme and a scheduling policy able to force desired equilibria conditions in networks. Specifically, we propose a repartition between uplink and downlink resources that is distributed among all the WiFi stations. We propose some extensions to standard DCF, in order to i) estimate the network status, in terms of per-station application requirements and channel access probability, and ii) emulate an access scheme based on best response strategies and AP mechanism design. The effectiveness of our solutions for controlling the resource sharing in WiFi networks is shown in various network scenarios.

We also consider the distributed resource allocation problem for random ad-hoc networks based on CSMA/CA access protocols, in order to suggest easily implementable solutions maximizing the throughput and minimizing the transmission delays in the network. To improve the CSMA protocol performances in a multi-hop scenario, we introduce a node synchronization mechanism, to reserve temporal slots to be used exclusively by a subset of nodes. We adopt a map coloring approach, revisiting some of the existing algorithms, and we study their effects to the a real traffic model via numerical simulations.

Chris Hadjicostis Distributed Weight Balancing in Directed Graphs.

We consider distributed systems whose components (nodes) can exchange information via interconnections (links) that form an arbitrary communication topology (graph). For many distributed control tasks that arise in such contexts, ranging from formation control and distributed averaging to consensus and distributed optimization, it is imperative to obtain a weight assignment on the links, such that the resulting graph is balanced. A weighted directed graph is balanced if, for each node, the sum of the weights on its incoming links is equal to the sum of the weights of its outgoing links. Distributed methodologies for obtaining weight assignments that balance undirected graphs are rather trivial; however, the task is significantly more challenging for directed graphs and has recently started to draw the attention of the research community. We present a class of distributed algorithms for asymptotically obtaining weight balance in directed graphs. The basic algorithm requires each node to assess its own imbalance (i.e., the difference of the sum of the weights on its incoming links minus the sum of the weights of its outgoing links) and, if this imbalance is positive, to (e.g., uniformly) increase the weights of its outgoing links so that it becomes balanced; a node is not allowed to modify the weights on its incoming links and does not take any action if its imbalance is negative. We analyze variations of this basic algorithm and obtain convergence bounds on their asymptotic performance.

Karl Hedrick Prey Modeling in Predatory/Prey Interaction – Risk avoidance, Group Foraging, and Communication.

In the animal kingdom there are many examples of teams that are highly efficient, robust, as well as heterogeneous and distributive by nature. The goal of the Heterogeneous Unmanned Networked Team (HUNT) project is to investigate the fundamental mechanism behind these teams and to explore their potential in collaborative control for autonomous agent teams. One of our focuses examines the prey side of the predator-prey interaction. Specifically, we examine the trade-off between foraging gain and predation risk faced by prey that is very similar to the trade-off faced by an autonomous agent team in a risky environment. In this talk we will present several bio-inspired movement rules for multi-agent information gathering and risk avoidance, which are based on the Domain of Danger concept used in animal behavior studies. Another focus of our work is the use of communication in animals. Teams of animals make use of many different types of communication such as visual, auditory, and chemical. Some animals also cooperate at times without any explicit communication. The different types of communication used will be examined for different types of tasks in the animal kingdom. These different types of communication could be useful in autonomous agent teams when dealing with loses in wireless communication due to a variety of expected or unexpected factors.

Steve Low Multi-period Optimal Procurement and Demand Response in the Presence of Uncertain Supply.

We propose a simple model that integrates two-period electricity markets, uncertainty in renewable generation, and real-time dynamic demand response. A load serving entity decides its day-ahead procurement to optimize expected social welfare a day before energy delivery. At delivery time when renewable generation is realized, it sets prices to manage demand and purchase additional power on the real-time market, if necessary, to balance supply and demand. We derive the optimal day-ahead decision, propose real-time demand response algorithm, and study the effect of volume and variability of renewable generation on these optimal decisions and on social welfare.

Sonia Martínez Distributed Constrained Optimization under Time-Varying Multi-agent Interactions.

Recent advances in communications, sensing and computation have made possible the deployment of mobile sensor networks for remote exploration and monitoring tasks. Relevant problems such as node reconfiguration and cooperative signal processing can be formulated as the constrained optimization of a simple sum of local functions. However, each local function may depend on a (possibly large) information set which is spread over the group. Thus, the absence of a centralized authority makes the collective task of deciding on a globally optimal solution challenging. In this talk, we present our recent progress in the analysis and design of distributed constrained optimization algorithms for a number of convex/non-convex problems. The algorithms have guaranteed convergence and can be implemented by agents subject to unpredictable link failures.

Mehran Mesbahi Viewing Networks as Systems.

I will explore certain results and observations pertaining to viewing networks as dynamic systems with inputs and outputs, thus placing them in the realm of control theory. Along the way, I will discuss the utility of adopting such a point of view, such as means of reasoning about the security of diffusion-based networks, identification of networks, the role of adaptation and randomness, as well as connections to certain problems in robotic, quantum, and biological networks.

Sean Meyn Mean-field Games for Fun and Profit.

Mean field games have emerged as a general approximation technique for applications in multi-agent models found in economics and control. This talk will provide an overview of recent results for both control and estimation. In the context of estimation, an extension of the particle filter for approximate nonlinear filtering is obtained as the solution to a mean field game. In the context of control, it is found that the approximating model can provide insight: In one application to coupled oscillators, the classical Kuramoto control law emerges as the approximate solution to a mean field game. A mean field model can also serve as a guide for basis selection for approximate dynamic programming algorithms, such as required in Q-learning.

Amit K. Roy-Chowdhury Integrated Sensing and Analysis in Distributed Camera Networks.

Over the past decade, large-scale camera networks have become increasingly prevalent in a wide range of applications, such as security and surveillance, disaster response, and environmental modeling. In many applications, bandwidth constraints, security concerns, and difficulty in storing and analyzing large amounts of data centrally at a single location necessitate the development of distributed camera network architectures. Thus, the development of distributed scene-analysis algorithms has received much attention lately. However, the performance of these algorithms often suffers because of the inability to effectively acquire the desired images, especially when the targets are dispersed over a wide field of view (FOV). In this talk, I will present our recent work on integrated sensing and analysis in a distributed camera network so as to maximize various scene-understanding performance criteria (e.g., tracking accuracy, best shot, and image resolution). We will show how the existing work in autonomous multiagent systems can be leveraged for this purpose; more specifically, game theory-based distributed optimization algorithms for dynamic camera network reconfiguration and consensus algorithms for scene analysis. An experimental test bed for evaluating such work will be described and some results from a real-life camera network presented.

Bruno Sinopoli Asymptotic Performance of Distributed Detection over Random Networks.

We consider the problem of distributed detection over random networks where sensors cooperate in order to decide between the two hypotheses on the state of nature. Specifically, we consider the running consensus distributed detector, where each sensor, at each time, improves its decision variable two-fold: 1) by averaging its decision variable with neighbors; and 2) by accounting on-the-fly for its new observation. The advantage of the described detection scheme with respect to the conventional consensus based detection is that it incorporates new observations in real time, at the same time scale as the averaging step. Another advantage is the absence of a fusion node, hence avoiding the bottleneck effect and the single point of failure effect. Distributed detection exhibits the tradeoff between the detection performance and the network connectivity, that is, the amount of utilized inter-sensor communications. In this work, we explicitly quantify the described tradeoff, and we reveal interesting detection behavior with respect to the network connectivity. Namely, we study the large deviations performance, i.e., the exponential decay rate of the detection error probability. We show that distributed detection exhibits a “phase change” behavior. When a certain measure of the network connectivity, the speed of consensus is above a threshold, then distributed detection is asymptotically equivalent to the optimal centralized detection, i.e., the error decay rate for distributed detection equals the Chernoff information. When the speed of consensus is below a threshold, then distributed detection achieves only a fraction of the Chernoff information rate; we quantify this achievable rate as a function of the speed of consensus. Simulation examples demonstrate our theoretical findings on the behavior of distributed detection over random networks.

Gaurav S. Sukhatme Planning and Decision-making for Underwater Robot Teams: Algorithms and Experiments.

We describe two ’networked robot’ problems in the underwater domain, both subject to the constraints of underwater acoustic communication. In the first, multiple robots are required to search for a target subject. We cast this as a distributed planning problem and give an algorithm for distributed coordination with limited communication. In the second, we describe and solve a problem where a single robot plans an optimal path based on data from multiple static underwater sensors. We conclude with a discussion of experimental results from a recent large-scale field study to illustrate the gap between algorithmic investigations and currently deployable systems.