Multi-agent planning

About: Multi-agent planning is a research topic. Over the lifetime, 288 publications have been published within this topic receiving 4128 citations.

Communication and interaction in multi-agent planning

Abstract: A method for synthesizing multi-agent plans from simpler single-agent plans is described. The idea is to insert communication acts into the single-agent plans so that agents can synchronize activities and avoid harmful interactions. Unlike most previous planning systems, actions are represented by sequences of states, rather than as simple state change operators. This allows the expression of more complex kinds of interaction than would otherwise be possible. An efficient method of interaction and safety analysis is then developed and used to identify critical regions in the plans. An essential feature of the method is that the analysis is performed without generating all possible interleavings of the plans, thus avoiding a combinatorial explosion. Finally, communication primitives are inserted into the plans and a supervisor process created to handle synchronization.

Formal models and algorithms for decentralized decision making under uncertainty

Abstract: Over the last 5 years, the AI community has shown considerable interest in decentralized control of multiple decision makers or "agents" under uncertainty. This problem arises in many application domains, such as multi-robot coordination, manufacturing, information gathering, and load balancing. Such problems must be treated as decentralized decision problems because each agent may have different partial information about the other agents and about the state of the world. It has been shown that these problems are significantly harder than their centralized counterparts, requiring new formal models and algorithms to be developed. Rapid progress in recent years has produced a number of different frameworks, complexity results, and planning algorithms. The objectives of this paper are to provide a comprehensive overview of these results, to compare and contrast the existing frameworks, and to provide a deeper understanding of their relationships with one another, their strengths, and their weaknesses. While we focus on cooperative systems, we do point out important connections with game-theoretic approaches. We analyze five different formal frameworks, three different optimal algorithms, as well as a series of approximation techniques. The paper provides interesting insights into the structure of decentralized problems, the expressiveness of the various models, and the relative advantages and limitations of the different solution techniques. A better understanding of these issues will facilitate further progress in the field and help resolve several open problems that we identify.

From one to many: planning for loosely coupled multi-agent systems

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 problem-independent, 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.

Communication and interaction in multi-agent planning

Abstract: A method for synthesizing multi-agent plans from simpler single-agent plans is described. The idea is to insert communication acts into the single-agent plans so that agents can synchronize activities and avoid harmful interactions. Unlike most previous planning systems, actions are represented by sequences of states, rather than as simple state change operators. This allows the expression of more complex kinds of interaction than would otherwise be possible. An efficient method of interaction and safety analysis is then developed and used to identify critical regions in the plans. An essential feature of the method is that the analysis is performed without generating all possible interleavings of the plans, thus avoiding a combinatorial explosion. Finally, communication primitives are inserted into the plans and a supervisor process created to handle synchronization.

Multi-agent planning as a dynamic search for social consensus

Abstract: When autonomous agents attempt to coordinate action, it is often necessary that they reach some kind of consensus. Reaching consensus has traditionally been dealt with in the Distributed Artificial Intelligence literature via negotiation. Another alternative is to have agents use a voting mechanism; each agent expresses its preferences, and a group choice mechanism is used to select the result. Some choice mechanisms are better than others, and ideally we would like one that cannot, be manipulated by untruthful agents. Coordination of actions by a group of agents corresponds to a group planning process. We here introduce a new multi-agent planning technique, that makes use of a dynamic, iterative search procedure. Through a process of group constraint aggregation, agents incrementally construct a plan that brings the group to a state maximizing social welfare. At each step, agents vote about the next joint action in the group plan (i.e., what the next transition state will be in the emerging plan) Using this technique agents need not fully reveal their preferences, and the set of alternative final states need not be generated in advance of a vote. With a minor variation, the entire procedure can be made resistant to untruthful agents.