arXiv subjects

Kai Cai

Publications and source records attributed to Kai Cai.

At least 55 records · Page 3Linked to original sources

arXiv2016

On Distributed Internal Model Principle for Output Regulation over Time-Varying Networks of Linear Heterogeneous Agents

We study a multi-agent output regulation problem, where not all agents have access to the exosystem's dynamics. We propose a distributed controller that solves the problem for linear, heterogeneous, and uncertain agent dynamics as well as time-varying directed networks. The distributed controller consists of two parts: (1) an exosystem generator that creates a local copy of the exosystem dynamics by using consensus protocols, and (2) a dynamic compensator that uses (again) consensus to approach the internal model of the exosystem and thereby achieves perfect output regulation. Our approach leverages methods from internal model based controller synthesis, multi-agent consensus over directed networks, and stability of time-varying linear systems; the derived result is an adaptation of the (centralized) internal model principle to the distributed, networked setting.

eess.SY↗

arXiv2016

Relative Coobservability in Decentralized Supervisory Control of Discrete-Event Systems

We study the new concept of relative coobservability in decentralized supervisory control of discrete-event systems under partial observation. This extends our previous work on relative observability from a centralized setup to a decentralized one. A fundamental concept in decentralized supervisory control is coobservability (and its several variations); this property is not, however, closed under set union, and hence there generally does not exist the supremal element. Our proposed relative coobservability, although stronger than coobservability, is algebraically well-behaved, and the supremal relatively coobservable sublanguage of a given language exists. We present an algorithm to compute this supremal sublanguage. Moreover, relative coobservability is weaker than conormality, which is also closed under set union; unlike conormality, relative coobservability imposes no constraint on disabling unobservable controllable events.

eess.SY↗

arXiv2015

On the Solvability of 3s/nt Sum-Network---A Region Decomposition and Weak Decentralized Code Method

We study the network coding problem of sum-networks with 3 sources and n terminals (3s/nt sum-network), for an arbitrary positive integer n, and derive a sufficient and necessary condition for the solvability of a family of so-called terminal-separable sum-network. Both the condition of terminal-separable and the solvability of a terminal-separable sum-network can be decided in polynomial time. Consequently, we give another necessary and sufficient condition, which yields a faster (O(|E|) time) algorithm than that of Shenvi and Dey ([18], (O(|E|^3) time), to determine the solvability of the 3s/3t sum-network. To obtain the results, we further develop the region decomposition method in [22], [23] and generalize the decentralized coding method in [21]. Our methods provide new efficient tools for multiple source multiple sink network coding problems.

cs.IT↗

arXiv2014

Delay-Robustness in Distributed Control of Timed Discrete-Event Systems Based on Supervisor Localization

Recently we studied communication delay in distributed control of untimed discrete-event systems based on supervisor localization. We proposed a property called delay-robustness: the overall system behavior controlled by distributed controllers with communication delay is logically equivalent to its delay-free counterpart. In this paper we extend our previous work to timed discrete-event systems, in which communication delays are counted by a special clock event {\it tick}. First, we propose a timed channel model and define timed delay-robustness; for the latter, a polynomial verification procedure is presented. Next, if the delay-robust property does not hold, we introduce bounded delay-robustness, and present an algorithm to compute the maximal delay bound (measured by number of ticks) for transmitting a channeled event. Finally, we demonstrate delay-robustness on the example of an under-load tap-changing transformer.

eess.SY↗

arXiv2014

Distributed Supervisory Control of Discrete-Event Systems with Communication Delay

This paper identifies a property of delay-robustness in distributed supervisory control of discrete-event systems (DES) with communication delays. In previous work a distributed supervisory control problem has been investigated on the assumption that inter-agent communications take place with negligible delay. From an applications viewpoint it is desirable to relax this constraint and identify communicating distributed controllers which are delay-robust, namely logically equivalent to their delay-free counterparts. For this we introduce inter-agent channels modeled as 2-state automata, compute the overall system behavior, and present an effective computational test for delay-robustness. From the test it typically results that the given delay-free distributed control is delay-robust with respect to certain communicated events, but not for all, thus distinguishing events which are not delay-critical from those that are. The approach is illustrated by a workcell model with three communicating agents.

eess.SY↗

arXiv2014

Relative Observability of Discrete-Event Systems and its Supremal Sublanguages

We identify a new observability concept, called relative observability, in supervisory control of discrete-event systems under partial observation. A fixed, ambient language is given, relative to which observability is tested. Relative observability is stronger than observability, but enjoys the important property that it is preserved under set union; hence there exists the supremal relatively observable sublanguage of a given language. Relative observability is weaker than normality, and thus yields, when combined with controllability, a generally larger controlled behavior; in particular, no constraint is imposed that only observable controllable events may be disabled. We design algorithms which compute the supremal relatively observable (and controllable) sublanguage of a given language, which is generally larger than the normal counterparts. We demonstrate the new observability concept and algorithms with a Guideway and an AGV example.

eess.SY↗

arXiv2014

Network Coding for $3$s$/n$t Sum-Networks

A sum-network is a directed acyclic network where each source independently generates one symbol from a given field $\mathbb F$ and each terminal wants to receive the sum $($over $\mathbb F)$ of the source symbols. For sum-networks with two sources or two terminals, the solvability is characterized by the connection condition of each source-terminal pair [3]. A necessary and sufficient condition for the solvability of the $3$-source $3$-terminal $(3$s$/3$t$)$ sum-networks was given by Shenvi and Dey [6]. However, the general case of arbitrary sources/sinks is still open. In this paper, we investigate the sum-network with three sources and $n$ sinks using a region decomposition method. A sufficient and necessary condition is established for a class of $3$s$/n$t sum-networks. As a direct application of this result, a necessary and sufficient condition of solvability is obtained for the special case of $3$s$/3$t sum-networks.

cs.IT↗

arXiv2013

Local Average Consensus in Distributed Measurement of Spatial-Temporal Varying Parameters: 1D Case

We study a new variant of consensus problems, termed `local average consensus', in networks of agents. We consider the task of using sensor networks to perform distributed measurement of a parameter which has both spatial (in this paper 1D) and temporal variations. Our idea is to maintain potentially useful local information regarding spatial variation, as contrasted with reaching a single, global consensus, as well as to mitigate the effect of measurement errors. We employ two schemes for computation of local average consensus: exponential weighting and uniform finite window. In both schemes, we design local average consensus algorithms to address first the case where the measured parameter has spatial variation but is constant in time, and then the case where the measured parameter has both spatial and temporal variations. Our designed algorithms are distributed, in that information is exchanged only among neighbors. Moreover, we analyze both spatial and temporal frequency responses and noise propagation associated with the algorithms. The tradeoffs of using local consensus, as compared to standard global consensus, include higher memory requirement and degraded noise performance. Arbitrary updating weights and random spacing between sensors are analyzed in the proposed algorithms.

eess.SY↗

arXiv2013

Supervisor Localization of Discrete-Event Systems based on State Tree Structures

Recently we developed supervisor localization, a top-down approach to distributed control of discrete-event systems in the Ramadge-Wonham supervisory control framework. Its essence is the decomposition of monolithic (global) control action into local control strategies for the individual agents. In this paper, we establish a counterpart supervisor localization theory in the framework of State Tree Structures, known to be efficient for control design of very large systems. In the new framework, we introduce the new concepts of local state tracker, local control function, and state-based local-global control equivalence. As before, we prove that the collective localized control behavior is identical to the monolithic optimal (i.e. maximally permissive) and nonblocking controlled behavior. In addition, we propose a new and more efficient localization algorithm which exploits BDD computation. Finally we demonstrate our localization approach on a model for a complex semiconductor manufacturing system.

eess.SY↗

arXiv2013

Average Consensus on Arbitrary Strongly Connected Digraphs with Time-Varying Topologies

We have recently proposed a "surplus-based" algorithm which solves the multi-agent average consensus problem on general strongly connected and static digraphs. The essence of that algorithm is to employ an additional variable to keep track of the state changes of each agent, thereby achieving averaging even though the state sum is not preserved. In this note, we extend this approach to the more interesting and challenging case of time-varying topologies: An extended surplus-based averaging algorithm is designed, under which a necessary and sufficient graphical condition is derived that guarantees state averaging. The derived condition requires only that the digraphs be arbitrary strongly connected in a \emph{joint} sense, and does not impose "balanced" or "symmetric" properties on the network topology, which is therefore more general than those previously reported in the literature.

eess.SY↗

arXiv2013

Supervision Localization of Timed Discrete-Event Systems

We study supervisor localization for real-time discrete-event systems (DES) in the Brandin-Wonham framework of timed supervisory control. We view a real-time DES as comprised of asynchronous agents which are coupled through imposed logical and temporal specifications; the essence of supervisor localization is the decomposition of monolithic (global) control action into local control strategies for these individual agents. This study extends our previous work on supervisor localization for untimed DES, in that monolithic timed control action typically includes not only disabling action as in the untimed case, but also ``clock preempting'' action which enforces prescribed temporal behavior. The latter action is executed by a class of special events, called ``forcible'' events; accordingly, we localize monolithic preemptive action with respect to these events. We demonstrate the new features of timed supervisor localization with a manufacturing cell case study, and discuss a distributed control implementation.

eess.SY↗

arXiv2013

Encoding Complexity of Network Coding with Two Simple Multicast Sessions

The encoding complexity of network coding for single multicast networks has been intensively studied from several aspects: e.g., the time complexity, the required number of encoding links, and the required field size for a linear code solution. However, these issues as well as the solvability are less understood for networks with multiple multicast sessions. Recently, Wang and Shroff showed that the solvability of networks with two unit-rate multicast sessions (2-URMS) can be decided in polynomial time. In this paper, we prove that for the 2-URMS networks: $1)$ the solvability can be determined with time $O(|E|)$; $2)$ a solution can be constructed with time $O(|E|)$; $3)$ an optimal solution can be obtained in polynomial time; $4)$ the number of encoding links required to achieve a solution is upper-bounded by $\max\{3,2N-2\}$; and $5)$ the field size required to achieve a linear solution is upper-bounded by $\max\{2,\lfloor\sqrt{2N-7/4}+1/2\rfloor\}$, where $|E|$ is the number of links and $N$ is the number of sinks of the underlying network. Both bounds are shown to be tight.

cs.IT↗

arXiv2012

Average Consensus on General Strongly Connected Digraphs

We study the average consensus problem of multi-agent systems for general network topologies with unidirectional information flow. We propose two (linear) distributed algorithms, deterministic and gossip, respectively for the cases where the inter-agent communication is synchronous and asynchronous. Our contribution is that in both cases, the developed algorithms guarantee state averaging on arbitrary strongly connected digraphs; in particular, this graphical condition does not require that the network be balanced or symmetric, thereby extending many previous results in the literature. The key novelty of our approach is to augment an additional variable for each agent, called "surplus", whose function is to locally record individual state updates. For convergence analysis, we employ graph-theoretic and nonnegative matrix tools, with the eigenvalue perturbation theory playing a crucial role.

eess.SY↗

arXiv2011

Convergence Time Analysis of Quantized Gossip Consensus on Digraphs

We have recently proposed quantized gossip algorithms which solve the consensus and averaging problems on directed graphs with the least restrictive connectivity requirements. In this paper we study the convergence time of these algorithms. To this end, we investigate the shrinking time of the smallest interval that contains all states for the consensus algorithm, and the decay time of a suitable Lyapunov function for the averaging algorithm. The investigation leads us to characterizing the convergence time by the hitting time in certain special Markov chains. We simplify the structures of state transition by considering the special case of complete graphs, where every edge can be activated with an equal probability, and derive polynomial upper bounds on convergence time.

eess.SY↗

arXiv2010

Autocorrelation-Run Formula for Binary Sequences

The autocorrelation function and the run structure are two basic notions for binary sequences, and have been used as two independent postulates to test randomness of binary sequences ever since Golomb 1955. In this paper, we prove for binary sequence that the autocorrelation function is in fact completely determined by its run structure.

cs.IT↗

arXiv2010

Confronting 3 Dimensional Time-dependent Jet Simulations with HST Observations

We perform state-of-the-art, 3D, time-dependent simulations of magnetized disk winds, carried out to simulation scales of 60 Astronomical Units, in order to confront optical HST observations of protostellar jets. We ``observe'' the optical forbidden line emission produced by shocks within our simulated jets and compare these with actual observations. Our simulations reproduce the rich structure of time varying jets, including jet rotation far from the source, an inner (up to 400 km/s) and outer (less than 100 km/s) component of the jet, and jet widths of up to 20 Astronomical Units in agreement with observed jets. These simulations when compared with the data are able to constrain disk wind models. In particular, models featuring a disk magnetic field with a modest radial spatial variation across the disk are favored.

astro-ph.SR↗

arXiv2010

Giant Planet Formation by Disk Instability: A Comparison Simulation With An Improved Radiative Scheme

There has been disagreement currently about whether cooling in protoplanetary disks can be sufficiently fast to induce the formation of gas giant protoplanets via gravitational instabilities. Simulations by our own group and others indicate that this method of planet formation does not work for disks around young, low- mass stars inside several tens of AU, while simulations by other groups show fragmentation into protoplanetary clumps in this region. To allow direct comparison in hopes of isolating the cause of the differences, we here present a high resolution three-dimensional hydrodynamics simulation of a protoplanetary disk, where the disk model, initial perturbation, and simulation conditions are essentially identical to those used in a set of simulations by Boss. As in earlier papers by the same author, Boss (2007, hereafter B07) purports to show that cooling is fast enough to produce protoplanetary clumps. Here, we evolve the same B07 disk using an improved version of one of our own radiative schemes and find that the disk does not fragment in our code but instead quickly settles into a state with only low amplitude nonaxisymmetric structure, which persists for at least several outer disk rotations. We see no rapid radiative or convective cooling. We conclude that the differences in results are due to different treatments of regions at and above the disk photosphere, and we explain at least one way in which the scheme in B07 may lead to artificially fast cooling.

astro-ph.SR↗

arXiv2010

A Grey-Box Approach to Automated Mechanism Design

Auctions play an important role in electronic commerce, and have been used to solve problems in distributed computing. Automated approaches to designing effective auction mechanisms are helpful in reducing the burden of traditional game theoretic, analytic approaches and in searching through the large space of possible auction mechanisms. This paper presents an approach to automated mechanism design (AMD) in the domain of double auctions. We describe a novel parametrized space of double auctions, and then introduce an evolutionary search method that searches this space of parameters. The approach evaluates auction mechanisms using the framework of the TAC Market Design Game and relates the performance of the markets in that game to their constituent parts using reinforcement learning. Experiments show that the strongest mechanisms we found using this approach not only win the Market Design Game against known, strong opponents, but also exhibit desirable economic properties when they run in isolation.

cs.GT↗