## From One to Many: Planning for Loosely Coupled Multi-Agent Systems

Citations: | 36 - 6 self |

### BibTeX

@MISC{Brafman_fromone,

author = {Ronen I. Brafman},

title = {From One to Many: Planning for Loosely Coupled Multi-Agent Systems},

year = {}

}

### OpenURL

### Abstract

Loosely coupled multi-agent systems are perceived as easier to plan for because they require less coordination between agent sub-plans. In this paper we set out to formalize this intuition. We establish an upper bound on the complexity of multi-agent planning problems that depends exponentially on two parameters quantifying the level of agents ’ coupling, and on these parameters only. The first parameter is problemindependent, and it measures the inherent level of coupling within the system. The second is problem-specific and it has to do with the minmax number of action-commitments per agent required to solve the problem. Most importantly, the direct dependence on the number of agents, on the overall size of the problem, and on the length of the agents ’ plans, is only polynomial. This result is obtained using a new algorithmic methodology which we call “planning as CSP+planning”. We believe this to be one of the first formal results to both quantify the notion of agents ’ coupling, and to demonstrate a multi-agent planning algorithm that, for fixed coupling levels, scales polynomially with the size of the problem.

### Citations

1823 |
STRIPS: A new approach to the application of theorem proving to problem solving
- Fikes, Nilsson
- 1971
(Show Context)
Citation Context ...ion of planning problems and their complexity must occur within some formal model. In this work we consider a minimalistic state-transition model expressed via the STRIPS classical planning language (=-=Fikes & Nilsson 1971-=-), slightly extended to associate actions with agents. To capture the level of interaction between agents we define and exploit the agent interaction (di)graph in which two agents are connected if one... |

584 |
Constraint Processing
- Dechter
(Show Context)
Citation Context ...nding a choice of coordination points into a globally consistent plan that corresponds to a certain combination of constraint satisfaction and planning. In general, a constraint satisfaction problem (=-=Dechter 2003-=-) is defined via a set of variables, U = {ui} n i=1 , with respective domains {Di} n i=1 , and set of constraints {ci} m i=1 . Each constraint ci is associated with a subset of variables {ui1, . . . ,... |

262 | HTN Planning: Complexity and Expressivity - Erol, Hendler, et al. - 1994 |

104 | Planning under continuous time and resource uncertainty: A challenge for ai
- Bresina, Dearden, et al.
- 2002
(Show Context)
Citation Context ...d and unload actions, then between every two actions requiring coordination, there would be many internal move actions. Another example would be the Rover domain that model NASA’s exploration rovers (=-=Bresina et al. 2002-=-). Imagine a set of rovers that explore a particular region. The public actions would be actions that carry out an experiment at a location, such as taking a measurement or a photo. These actions are ... |

82 | Graph searching and a min-max theorem for tree-width
- Seymour, Thomas
- 1993
(Show Context)
Citation Context ...s, and there is an edge between ui and uj just if both participate in some constraint c. Informally, the treewidth of a graph is a measure of its ”cliquishness,” or how tightly coupled its nodes are (=-=Seymour & Thomas 1993-=-). For example, the tree-width of a tree is 1, regardless of its size, whereas the tree-width of a complete graph over n nodes is n. Let δ denote the minimal coordination-sequence length under which a... |

78 | Distributed Problem Solving and Planning
- Durfee
- 1999
(Show Context)
Citation Context ...et the agents ”fill-in the details” on their own. In fact, this intuitive principle is already adopted one way or another in many domainspecific multi-agent (and, in particular, multi-robot) systems (=-=Durfee 1999-=-). Likewise, this principle lies in the heart of (both domain-specific and general-purpose) hierarchical planning systems (e.g., (Erol, Hendler, & Nao 1994; Knoblock 1994; Clement, Durfee, & Barrett 2... |

62 | Branching and pruning: an optimal temporal POCL planner based on constraint programming - Vidal, Geffner |

56 | A multiagent planning architecture
- Wilkins, Myers
- 1998
(Show Context)
Citation Context ...ombine their expertise to solve problems that are beyond their individual capabilities.” (Durfee 1999). A nice example of this approach in the context of planning and scheduling has been proposed in (=-=Wilkins & Myers 1998-=-), where sophisticated systems for planning and scheduling are decomposed into modules, each of which is transformed into an agent, allowing experimentation with different degrees of coupling between ... |

44 | Factored planning: How, when, and when not
- Brafman, Domshlak
- 2006
(Show Context)
Citation Context ...pproaches), or exponential in the length of the shortest plan (for BFS-style procedures). The exceptions would be only some recently-proposed algorithms for factored planning (Amir & Engelhardt 2003; =-=Brafman & Domshlak 2006-=-; Kelareva et al. 2007) that we build upon in our work here. The MA-STRIPS solving framework we propose here combines some technical ideas underlying two such factored planning algorithms of (Brafman ... |

38 | Factored planning
- Amir, Engelhardt
- 2003
(Show Context)
Citation Context ...andard planning-as-CSP approaches), or exponential in the length of the shortest plan (for BFS-style procedures). The exceptions would be only some recently-proposed algorithms for factored planning (=-=Amir & Engelhardt 2003-=-; Brafman & Domshlak 2006; Kelareva et al. 2007) that we build upon in our work here. The MA-STRIPS solving framework we propose here combines some technical ideas underlying two such factored plannin... |

37 | Complexity results for standard benchmark domains in planning - Helmert |

23 | Multi-entity models
- Moses, Tennenholtz
- 1995
(Show Context)
Citation Context ... problems expressible in a minimalistic MA-extension of the STRIPS language (Fikes & Nilsson 1971). In particular, the problems considered here comprise the seminal automata-based multientity models (=-=Moses & Tennenholtz 1995-=-). In what follows, we formalize this extension of STRIPS, as well as some of its useful derivatives that we then employ in the problemsolving part of the paper. Definition 1 An MA-STRIPS problem for ... |

18 | Abstract reasoning for planning and coordination - Clement, Durfee, et al. - 2007 |

18 |
Distributed Problem Solving and Planning," in Multi-Agent Systems and Applications
- Durfee
- 2006
(Show Context)
Citation Context ...et the agents “fill-in the details” on their own. In fact, this intuitive principle is already adopted one way or another in many domainspecific multi-agent (and, in particular, multi-robot) systems (=-=Durfee 1999-=-). Likewise, this principle lies in the heart of (both domain-specific and general-purpose) hierarchical planning systems (e.g., (Erol, Hendler, & Nao 1994; Knoblock 1994; Clement, Durfee, & Barrett 2... |

3 |
Automatically generating abstractions for planning. AIJ 68(2):243–302
- Knoblock
- 1994
(Show Context)
Citation Context ...rticular, multi-robot) systems (Durfee 1999). Likewise, this principle lies in the heart of (both domain-specific and general-purpose) hierarchical planning systems (e.g., (Erol, Hendler, & Nao 1994; =-=Knoblock 1994-=-; Clement, Durfee, & Barrett 2007)). Our objective here is to operationalize this principle in a generic, domain-independent manner in systems that do not necessarily exhibit substantial hierarchy amo... |