Corpus ID: 1122928

Probabilistic Planning via Determinization in Hindsight

@inproceedings{Yoon2008ProbabilisticPV,
  title={Probabilistic Planning via Determinization in Hindsight},
  author={S. Yoon and Alan Fern and R. Givan and S. Kambhampati},
  booktitle={AAAI},
  year={2008}
}
This paper investigates hindsight optimization as an approach for leveraging the significant advances in deterministic planning for action selection in probabilistic domains. Hindsight optimization is an online technique that evaluates the one-step-reachable states by sampling future outcomes to generate multiple non-stationary deterministic planning problems which can then be solved using search. Hindsight optimization has been successfully used in a number of online scheduling applications… Expand
Improving Determinization in Hindsight for On-line Probabilistic Planning
TLDR
Three significant improvements are presented to help determinization in hindsight to scale significantly better on large or mostly-deterministic problems. Expand
Generalizing the Role of Determinization in Probabilistic Planning
TLDR
This work shows that learning a good determinization for a planning domain can be done efficiently and can improve performance, and introduces a planner, FF-LAO*, that outperforms state-of-the-art probabilistic planners on several well-known competition benchmarks. Expand
Hindsight Optimization for Probabilistic Planning with Factored Actions
TLDR
A novel framework for on-line stochastic planning is proposed, by embedding the idea of hindsight optimization into a reduction to integer linear programming, which is general purpose by working with domain specifications over factored state and action spaces, and is scalable in principle to exponentially large action spaces. Expand
Anytime optimal MDP planning with trial-based heuristic tree search
TLDR
The main contribution of this thesis, the Trial-based Heuristic Tree Search (THTS) framework, allows the description of algorithms in terms of only six ingredients that can be mixed and matched at will, and shows theoretically and empirically that all proposed algorithms improve significantly over the application of the state-of-the-art approach. Expand
World Planning via Hindsight Optimization
In classical planning, it is assumed that all relevant aspects of the world are known at planning time. In many domains, however, the existence of certain objects or the states of certain fluents areExpand
Planning Under Temporal Uncertainty Using Hindsight Optimization
A robot task planner must be able to tolerate uncertainty in the durations of commanded actions and uncertainty in the time of occurrence of exogenous events. Sophisticated temporal reasoningExpand
World Planning for Robots via Hindsight Optimization
Classical planning makes the closed world assumption in which all relevant aspects of the world are known at planning time. While this assumption holds in some domains, in many practical roboticsExpand
Continual On-line Planning as Decision-Theoretic Incremental Heuristic Search
TLDR
An extension of Russell and Wefald’s decision-theoretic algorithm that can cope with the inadmissible heuristic search is proposed, and the algorithm, DTOCS, handles the complexities of the online setting by balancing deliberative planning and real-time response. Expand
Integrating Paradigms for Approximate Probabilistic Planning
TLDR
RETRASE is introduced, a novel MDP solver that derives approximate policies by extracting problem structure and learning its parameters under heuristic guidance and several extensions of RETRASE are outlined and new directions for unifying paradigms in probabilistic planning prompted by RETRase’s success are outlined. Expand
Incremental plan aggregation for generating policies in MDPs
TLDR
A way to generate policies in MDPs by determinizing the given MDP model into a classical planning problem, and using sequential Monte-Carlo simulations of the partial policies before execution, in order to assess the probability of replanning for a policy during execution is described. Expand
...
1
2
3
4
5
...

References

SHOWING 1-10 OF 21 REFERENCES
Sequential Monte Carlo in reachability heuristics for probabilistic planning
TLDR
This work applies heuristic search to conformant probabilistic planning by adapting planning graph heuristics developed for non-deterministic planning, and empirically demonstrates how this efficient, but sometimes suboptimal, approach enables the planner to solve much larger problems than an existing optimal bounded length Probabilistic planner and still find reasonable quality solutions. Expand
FF-Replan: A Baseline for Probabilistic Planning
TLDR
This paper gives the first technical description of FF-Replan and provides an analysis of its results on all of the recent IPPC-04 andIPPC-06 domains, in the hope that this will inspire extensions and insight into the approach and planning domains themselves that will soon lead to the dethroning of FF -Replan. Expand
Concurrent Probabilistic Planning in the Graphplan Framework
TLDR
This work presents an extension of the full Graphplan framework to probabilistic domains that demonstrates a method of efficiently finding optimal contingency plans using a goal regression search, and produces a planner that is competitive with the state of the art. Expand
FF + FPG: Guiding a Policy-Gradient Planner
TLDR
This paper shows how to use an external teacher to guide FPG's exploration using the actions suggested by FF's heuristic (Hoffmann 2001), as FF-replan has proved efficient for probabilistic re-planning. Expand
Policy Generation for Continuous-time Stochastic Domains with Concurrency
TLDR
This work adopts the framework of Younes, Musliner, & Simmons for planning with concurrency in continuous-time stochastic domains and implements a set of concrete techniques for policy generation, failure analysis, and repair in TEMPASTIC, a novel temporal probabilistic planner. Expand
PEGASUS: A policy search method for large MDPs and POMDPs
We propose a new approach to the problem of searching a space of policies for a Markov decision process (MDP) or a partially observable Markov decision process (POMDP), given a model. Our approach isExpand
The FF Planning System: Fast Plan Generation Through Heuristic Search
TLDR
A novel search strategy is introduced that combines hill-climbing with systematic search, and it is shown how other powerful heuristic information can be extracted and used to prune the search space. Expand
From Conformant into Classical Planning: Efficient Translations that May Be Complete Too
TLDR
This paper considers an alternative, more powerful translation based on the introduction of epistemic tagged literals KL/t where L is a literal in P and t is a set of literals in P unknown in the initial situation, and shows that a conformant planner based on this translation solves some interesting domains that cannot be solved by other planners. Expand
A framework for simulation-based network control via hindsight optimization
  • E. Chong, R. Givan, H. Chang
  • Computer Science
  • Proceedings of the 39th IEEE Conference on Decision and Control (Cat. No.00CH37187)
  • 2000
TLDR
A novel approach for designing network control algorithms that incorporate traffic models that can be viewed as stochastic predictions about the future network state, and can be used to generate traces of potential future network behavior using a technique called hindsight optimization. Expand
Probabilistic planning vs replanning
TLDR
A baseline test for probabilistic interestingness is tested, along with some examples of its application, and an analysis of the latest probabilism competition problems is attempted, to suggest some improvements that could be made for future instances of the competition. Expand
...
1
2
3
...