Towards Synthesizing Optimal Coordination Modules for Distributed Agents
نویسندگان
چکیده
In a discrete-event framework, we define the concept of a coordinable language and show that it is the necessary and sufficient existence condition of coordination modules for distributed agents to achieve conformance to a given inter-agent constraint language. We also present a synthesis algorithm to compute near optimal coordination modules. Background & Notation We use the following notations from language and automata theory (Cassandras and Lafortune 1999): • For an event set Σ, Σ denotes the set of all finite strings over Σ, including the empty string ε; and for two event sets Σ ⊆ Σ, PΣ2,Σ1 denotes the natural projection from (Σ) to (Σ), erasing all σ ∈ Σ − Σ in s ∈ (Σ). • For a language L over an event set Σ, i.e., L ⊆ Σ, L̄ denotes the set of all prefixes of its strings. L is said to be prefix-closed if L = L̄. • For an automaton A = ( X,Σ, δ, x0 , X A m ) , L(A) andLm(A) denote its generated prefix-closed andmarked languages, respectively; and for x ∈ X, s ∈ (Σ), δ(s, x)! denotes that δ(s, x) is defined; Trim(A) denotes the procedure that computes and returns a nonblocking automatonwhich generates the samemarked language as A; and A = A1 ‖ A2 denotes that automaton A is the synchronous product of the two automata A1 and A2. Consider a system modeled by an automaton A, with the event set Σ partitioned into (i) Σ = Σc . ∪ Σuc and (ii) Σ = Σo . ∪ Σuo, where Σ A c , Σ A uc, Σ A o and Σ A uo denote the sets of controllable, uncontrollable, observable and unobservable events of A, respectively. Let K be a sublanguage of Lm(A), i.e.,K ⊆ Lm(A). The statements ‘K is controllable w.r.tA, Σc ’ and ‘K is observable w.r.tA, PΣA,ΣAo ’ refer, respectively, to the concepts of controllability (Ramadge and Wonham 1987) and observability (Lin and Wonham 1988) of a language K in supervisory control theory. Finally, for an automatonC, theSupcon(C,A,Σc ) procedure (Wonham and Ramadge 1987) computes a nonblocking automatonS such thatLm(S) is the supremal controllable subCopyright c © 2008, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. language (Ramadge andWonham 1987) of Lm(C)∩Lm(A) w.r.t A and Σc . Discrete-Event Agents & Coordination Consider a system of two agents modeled by the respective automata Ai = (X i,Σi , δi , xi 0 , X Ai m ) (i ∈ {1, 2}), where Σ1 ∩ Σ2 = ∅. The event set Σi of agent Ai is partitioned into the controllable set Σi c and the uncontrollable set Σi uc . In enabling distributed agents to coordinate, each agentAi is equipped with a coordination module (CM) modeled by an automaton Si with the following properties: 1. Σi = Σi ∪ ComSet(Si, Aj), where ComSet(Si, Aj) ⊆ Σj , (i, j ∈ {1, 2}, i 6= j). Σi is called the coordination event set for agent Ai, and ComSet(Si, Aj) is the set of events that agent Aj needs to communicate to Ai to synchronize Si. 2. Si is Σ Ai uc -enabling, namely, (∀s ∈ (Σ i))(∀σ ∈ Σi uc ) ((s ∈ L(Si ‖ Ai) and PΣSi ,Σi (s)σ ∈ L(Ai)) ⇒ (sσ ∈ L(Si ‖ Ai))). 3. Si and Sj are cooperative, namely, (∀s ∈ (Σ))(∀σ ∈ ComSet(Si, Aj)) ((PΣA,ΣAj (s)σ ∈ L(Aj) and PΣA,ΣSj (s)σ ∈ L(Sj)) ⇒ (PΣA,ΣSi (s)σ ∈ L(Si))). Let A = A1 ‖ A2 and S12 = (S1, S2) denote the CM pair of S1 and S2. Write S12/A for the system of two agents A1 and A2 coordinating through their respective CM’s. Definition 1. Coordinated Behaviors 1. Prefix-closed coordinated behavior L(S12/A) (a) ε ∈ L(S12/A). (b) (∀s ∈ L(S12/A))(∀σ ∈ Σi) (sσ ∈ L(S12/A) ⇔ (sσ ∈ L(A) and (PΣA,ΣSi (s)σ ∈ L(Si)))). 2. Marked coordinated behavior Lm(S12/A) Lm(S12/A) = L(S12/A)∩Lm(A)∩Lm(S1)∩Lm(S2). CM pair S12 is nonblocking if Lm(S12/A) = L(S12/A). Definition 2. Coordinable Language: Let Σcom ⊆ Σ. A languageK ⊆ Lm(A) is coordinable w.r.t A and Σcom if 1. K is controllable w.r.t A and Σc = Σ A1 c ∪ Σ A2 c ; and 2. K is observable w.r.t A and PΣA,ΣAi∪Σcom (i ∈ {1, 2}). Proceedings of the Twenty-Third AAAI Conference on Artificial Intelligence (2008)
منابع مشابه
Synthesizing Coordination Requirements for Heterogeneous Autonomous Agents
As agents move into ever more important applications, there is a natural growth in interest in techniques for synthesizing multiagent systems. We describe an approach for engineering the coordination requirements of a multiagent system based on an analysis of conversation instances extracted from usage scenarios. This approach exploits the notion of Dooley graphs that were recently introduced t...
متن کاملOptimal Distributed Generation Planning in Radial Distribution Networks Considering Protection Coordination Limits
Distributed generation (DG) has been widely used in distribution network to reduce the energy losses, improve voltage profile and system reliability, etc. The location and capacity of DG units can influence on probability of protection mal-operation in distribution networks. In this paper, a novel model for DG planning is proposed to find the optimum DG location and sizing in radial distr...
متن کاملOn-line coordination among discrete-event agents
This paper addresses a novel coordination problem for distributed agents in a discrete-event setting. We introduce and study a predicate coordination problem as the problem of distributed agents interacting and communicating between themselves to satisfy (the invariance of) a global predicate specifying an inter-agent constraint. We then develop an optimal coordination policy by which the agent...
متن کاملOn deconflicting local coordination among agents
For conflict resolution between local coordination modules of distributed agents, synthesized based on inter-agent constraints, two original ideas are proposed. The first is a designer-comprehensible Distributed Constraint Specification Network (DCSN) for describing the constraint relationships among agents. The second is an algorithm using cutset theory for generating, from a given DCSN, an AN...
متن کاملReputation-based Distributed Coordination for Heterogeneous Autonomous Agents Towards Effective Coordination, Cooperation & Coalition Formation for Autonomous Software Agents Belonging to Different End-Users
Decentralized, partially or fully distributed multi-agent coordination, cooperation and coalition formation have been studied extensively over the past 20 years. Ensembles of autonomous software, robotic or other autonomous agents may need to coordinate and form coalitions in order to share resources, divide-and-conquer tasks, or enable themselves to complete tasks too complex for any individua...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2008