Learning Planning Action Models with Numerical Information and Logic Relationships Using Classification Techniques

Segura-Muros, José Á.; Pérez, Raúl; Fernández-Olivares, Juan

doi:10.1007/978-3-030-00374-6_35

Cited by 1 publication

(2 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Most of them deals with partial observations (except Observer [29] and OLAM [30] that needs complete observations). Among these approaches are ARMS [5], Louga [31], Plan-Milner algorithm [12], AIA [32], [33] or FAMA [34]. Among these works, the ARMS system is the most known.…”

Section: Related Workmentioning

confidence: 99%

“…Therefore, we consider that domain accuracy, that we define as the capacity of a learned domain to solve planning problems that were not used in the training dataset, is a better performance indicator than syntactical distance in practice. Third, even if some approaches, e.g., [7], [12], [13] are able to learn from noisy and/or partially observable data, few approaches are able to handle very high levels of noise and high levels of partial observations as can be encountered in real world applications, especially in robotics where observations are extracted using miscalibrated or noisy sensors.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

An Accurate PDDL Domain Learning Algorithm from Partial and Noisy Observations

Grand

Pellier

Fiorino

2022

2022 IEEE 34th International Conference on Tools With Artificial Intelligence (ICTAI)

View full text Add to dashboard Cite

This paper presents a novel approach to learn PDDL domain called AMLSI (Action Model Learning with State machine Interaction) based on grammar induction. AMLSI learns with no prior knowledge from a training dataset made up of action sequences built by random walks and by observing state transitions. The domain learnt is accurate enough to be used without human proofreading in a planner even with very highly partial and noisy observations. Thus AMLSI takles a key issue for domain learning that is the ability to plan with the learned domains. It often happens that small learning errors lead to domains that are unusable for planning. AMLSI contribution is to learn domains from partial and noisy observations with sufficient accuracy to allow planners to solve new problems. Compared to other approaches, AMLSI uses smaller training datasets and exploits both feasible and infeasible generated action sequences.

show abstract