This paper is a report on research towards the development of an abstraction-based framework for decision-theoretic planning. We make use of two planning approaches in the context of probabilistic planning: planning by abstraction and planning graphs. To create abstraction hierarchies our planner uses an adapted version of a hierarchical planner under uncertainty, and to search for plans, we pr...

Due to the high complexity of probabilistic planning algorithms, roboticists often opt for deterministic replanning paradigms, which can quickly adapt the current plan to the environment’s changes. However, probabilistic planning suffers in practice from the common misconception that it is needed to generate complete or closed policies, which would not require to be adapted on-line. In this wor...

Probabilistic temporal planning attempts to find good policies for acting in domains with concurrent durative tasks, multiple uncertain outcomes, and limited resources. These domains are typically modelled as Markov decision problems and solved using dynamic programming methods. This paper demonstrates the application of reinforcement learning — in the form of a policy-gradient method — to thes...

We present an any-time concurrent probabilistic temporal planner that includes continuous and discrete uncertainties and metric functions. Our approach is a direct policy search that attempts to optimise a parameterised policy using gradient ascent. Low memory use, plus the use of function approximation methods, plus factorisation of the policy, allow us to scale to challenging domains. This Fa...

Classical arti cial intelligence planning techniques can operate in large domains but traditionally assume a deterministic universe. Operations research planning techniques can operate in probabilistic domains but break when the domains approach realistic sizes. maxplan is a new probabilistic planning technique that aims at combining the best of these two worlds. maxplan converts a planning ins...

In the probabilistic track of the IPC5 —the last International Planning Competitions— a probabilistic planner based on combining deterministic planning with replanning —FF-REPLAN—outperformed the other competitors. This probabilistic planning paradigm discarded the probabilistic information of the domain, just considering for each action its nominal effect as a deterministic effect. Thus, in ce...

In this paper, a new online robot motion planner is developed for systematically exploring unknown environ¬ments by intelligent mobile robots in real-time applications. The algorithm takes advantage of sensory data to find an obstacle-free start-to-goal path. It does so by online calculation of the Generalized Voronoi Graph (GVG) of the free space, and utilizing a combination of depth-first an...

Finding paths in high-dimensional gemetric spaces is a provably hard problem. Recently, a general randomized planning scheme has emerged as an e ective approach to solve this problem. In this scheme, the planner samples the space at random and build a network of simple paths, called a probabilistic roadmap. This paper describes a basic probabilistic roadmap planner, which is easily parallelizab...

This paper presents a probabilistic planner capable of nding paths for a exible surface patch. The planner is based on the Probabilistic Roadmap approach to path planning while the surface patch is modeled as a low degree B ezier surface. We assume that we are dealing with an elastic part and deene an approximate energy model for the part. The energy function penalizes excessive shear and bendi...