Results 1 -
9 of
9
Distributed Heuristic Forward Search for Multi-agent Planning
"... This paper deals with the problem of classical planning for multiple cooperative agents who have private information about their local state and capabilities they do not want to reveal. Two main approaches have recently been proposed to solve this type of prob-lem – one is based on reduction to dist ..."
Abstract
-
Cited by 3 (0 self)
- Add to MetaCart
(Show Context)
This paper deals with the problem of classical planning for multiple cooperative agents who have private information about their local state and capabilities they do not want to reveal. Two main approaches have recently been proposed to solve this type of prob-lem – one is based on reduction to distributed constraint satisfaction, and the other on partial-order planning techniques. In classical single-agent planning, constraint-based and partial-order planning techniques are currently dominated by heuristic forward search. The question arises whether it is possible to formulate a distributed heuristic forward search algorithm for privacy-preserving classical multi-agent planning. Our work provides a pos-itive answer to this question in the form of a general approach to distributed state-space search in which each agent performs only the part of the state expansion relevant to it. The resulting algorithms are simple and efficient – outperforming previous algorithms by orders of magnitude – while offering similar flexibility to that of forward-search algorithms for single-agent planning. Furthermore, one particular variant of our general approach yields a distributed version of the a * algorithm that is the first cost-optimal distributed algorithm for privacy-preserving planning. 1.
Decoupling the multiagent disjunctive temporal problem
- In Proceedings of the Twenty-Seventh Conference on Artificial Intelligence (AAAI-13
, 2013
"... The Multiagent Disjunctive Temporal Problem (MaDTP) is a general constraint-based formulation for scheduling problems that involve interdependent agents. Decoupling agents ’ interdependent scheduling problems, so that each agent can manage its schedule independently, requires agents to adopt additio ..."
Abstract
-
Cited by 2 (1 self)
- Add to MetaCart
The Multiagent Disjunctive Temporal Problem (MaDTP) is a general constraint-based formulation for scheduling problems that involve interdependent agents. Decoupling agents ’ interdependent scheduling problems, so that each agent can manage its schedule independently, requires agents to adopt additional local constraints that effec-tively subsume their interdependencies. In this paper, we present the first algorithm for decoupling MaDTPs. Our distributed algorithm is provably sound and com-plete. Our experiments show that the relative efficiency of using temporal decoupling to find solution spaces for MaDTPs, compared to algorithms that find complete solution spaces, improves with the interconnectedness be-tween agents schedules, leading to orders of magnitude relative speeedup. However, decoupling by its nature restricts agents ’ scheduling flexibility; we define novel flexibility metrics for MaDTPs, and show empirically how the flexibility sacrificed depends on the degree of coupling between agents ’ schedules.
A flexible approach to modeling unpredictable events in MDPs
- In Proc. of the International Conference on Automated Planning and Scheduling
, 2013
"... In planning with a Markov decision process (MDP) frame-work, there is the implicit assumption that the world is pre-dictable. Practitioners must simply take it on good faith the MDP they have constructed is comprehensive and accurate enough to model the exact probabilities with which all events may ..."
Abstract
-
Cited by 2 (2 self)
- Add to MetaCart
In planning with a Markov decision process (MDP) frame-work, there is the implicit assumption that the world is pre-dictable. Practitioners must simply take it on good faith the MDP they have constructed is comprehensive and accurate enough to model the exact probabilities with which all events may occur under all circumstances. Here, we challenge the conventional assumption of complete predictability, arguing that some events are inherently unpredictable. Towards more effectively modeling problems with unpredictable events, we develop a hybrid framework that explicitly distinguishes de-cision factors whose probabilities are not assigned precisely while still representing known probability components using conventional principled MDP transitions. Our approach is also flexible, resulting in a factored model of variable abstraction whose usage for planning results in different levels of approxi-mation. We illustrate the application of our framework to an intelligent surveillance planning domain. 1
Decoupling the Multiagent Disjunctive Temporal Problem
"... Abstract The Multiagent Disjunctive Temporal Problem (MaDTP) is a general constraint-based formulation for scheduling problems that involve interdependent agents. Decoupling agents' interdependent scheduling problems, so that each agent can manage its schedule independently, requires agents to ..."
Abstract
- Add to MetaCart
Abstract The Multiagent Disjunctive Temporal Problem (MaDTP) is a general constraint-based formulation for scheduling problems that involve interdependent agents. Decoupling agents' interdependent scheduling problems, so that each agent can manage its schedule independently, requires agents to adopt additional local constraints that effectively subsume their interdependencies. In this paper, we present the first algorithm for decoupling MaDTPs. Our distributed algorithm is provably sound and complete. Our experiments show that the relative efficiency of using temporal decoupling to find solution spaces for MaDTPs, compared to algorithms that find complete solution spaces, improves with the interconnectedness between agents schedules, leading to orders of magnitude relative speeedup. However, decoupling by its nature restricts agents' scheduling flexibility; we define novel flexibility metrics for MaDTPs, and show empirically how the flexibility sacrificed depends on the degree of coupling between agents' schedules.
under Uncertainty
, 2013
"... model background: FSPC Factored FSPC practical factored FFSPC results ..."
Robot Planning under Uncertainty with Unpredictable Events
"... Abstract. In planning robot behavior with a Markov decision process (MDP) framework, there is the implicit assumption that the world is predictable. Practitioners must simply take it on good faith the MDP they have constructed is comprehensive and accurate enough to model the exact probabilities wit ..."
Abstract
- Add to MetaCart
(Show Context)
Abstract. In planning robot behavior with a Markov decision process (MDP) framework, there is the implicit assumption that the world is predictable. Practitioners must simply take it on good faith the MDP they have constructed is comprehensive and accurate enough to model the exact probabilities with which all events may occur under all circum-stances that the robot may encounter. Here, we challenge the conventional assumption of complete predictability, arguing that some events are in-herently unpredictable. Towards more effectively modeling problems with unpredictable events, we develop a flexible framework that explicitly dis-tinguishes decision factors whose probabilities are not assigned precisely while still representing known probability components using conventional principled MDP transitions. Our approach is also flexible, resulting in a factored model of variable abstraction whose usage for planning results in different levels of approximation. We illustrate the usage of our modeling framework in a robot surveillance domain. 1
Exploiting Separability in Multiagent Planning with
"... Recent years have seen significant advances in techniques for op-timally solving multiagent problems represented as decentralized partially observable Markov decision processes (Dec-POMDPs). A new method achieves scalability gains by converting Dec-POMDPs into continuous state MDPs. This method reli ..."
Abstract
- Add to MetaCart
(Show Context)
Recent years have seen significant advances in techniques for op-timally solving multiagent problems represented as decentralized partially observable Markov decision processes (Dec-POMDPs). A new method achieves scalability gains by converting Dec-POMDPs into continuous state MDPs. This method relies on the assumption of a centralized planning phase that generates a set of decentralized policies for the agents to execute. However, scalability remains limited when the number of agents or problem variables becomes large. In this paper, we show that, under certain separability condi-tions of the optimal value function, the scalability of this approach can increase considerably. This separability is present when there is locality of interaction, which — as other approaches (such as those based on the ND-POMDP subclass) have already shown — can be exploited to improve performance. Unlike most previous meth-ods, the novel continuous-state MDP algorithm retains optimality and convergence guarantees. Results show that the extension us-ing separability can scale to a large number of agents and domain variables while maintaining optimality.
ABSTRACT a
"... Recent years have seen significant advances in techniques for optimally solving multiagent problems represented as decentralized partially observable Markov decision processes (Dec-POMDPs). A new method achieves scalability gains by converting Dec-POMDPs into continuous state MDPs. This method relie ..."
Abstract
- Add to MetaCart
(Show Context)
Recent years have seen significant advances in techniques for optimally solving multiagent problems represented as decentralized partially observable Markov decision processes (Dec-POMDPs). A new method achieves scalability gains by converting Dec-POMDPs into continuous state MDPs. This method relies on the assumption of a centralized planning phase that generates a set of decentralized policies for the agents to execute. However, scalability remains limited when the number of agents or problem variables becomes large. In this paper, we show that, under certain separability conditions of the optimal value function, the scalability of this approach can increase considerably. This separability is present when there is locality of interaction, which — as other approaches (such as those based on the ND-POMDP subclass) have already shown — can be exploited to improve performance. Unlike most previous methods, the novel continuous-state MDP algorithm retains optimality and convergence guarantees. Results show that the extension using separability can scale to a large number of agents and domain variables while maintaining optimality.
