Pasar al contenido principal

APRENDIZAJE PARA PLANIFICACION SENSIBLE AL HUMANO

PID2021-127647NB-C22

•
Nombre agencia financiadora Agencia Estatal de Investigación
Acrónimo agencia financiadora AEI
Programa Programa Estatal para Impulsar la Investigación Científico-Técnica y su Transferencia
Subprograma Subprograma Estatal de Generación de Conocimiento
Convocatoria Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Año convocatoria 2021
Plan Estatal 2021-2023 Plan Estatal de Investigación Científica y Técnica y de Innovación
Centro beneficiario UNIVERSITAT POLITÈCNICA DE VALÈNCIA

Resultados relacionados

Resultados totales (Incluyendo duplicados): 13
Encontrada(s) 1 página(s)

Repositorio Digital de la UPF
Publicación Artículo científico (JournalArticle)

Generalized planning as heuristic search: a new planning search-space that leverages pointers over objects

  • Segovia-Aguas, Javier
  • Jiménez, Sergio
  • Jonsson, Anders, 1973-
Planning as heuristic search is one of the most successful approaches to classical planning but unfortunately, it does not trivially extend to Generalized Planning (GP); GP aims to compute algorithmic solutions that are valid for a set of classical planning instances from a given domain, even if these instances differ in their number of objects, the initial and goal configuration of these objects and hence, in the number (and possible values) of the state variables. State-space search, as it is implemented by heuristic planners, becomes then impractical for GP. In this paper we adapt the planning as heuristic search paradigm to the generalization requirements of GP, and present the first native heuristic search approach to GP. First, the paper introduces a new pointer-based solution space for GP that is independent of the number of classical planning instances in a GP problem and the size of those instances (i.e. the number of objects, state variables and their domain sizes). Second, the paper defines an upgraded version of our GP algorithm, called Best-First Generalized Planning (BFGP), that implements a best-first search in our pointer-based solution space for GP. Lastly, the paper defines a set of evaluation and heuristic functions for BFGP that assess the structural complexity of the candidate GP solutions, as well as their fitness to a given input set of classical planning instances. The computation of these evaluation and heuristic functions does not require grounding states or actions in advance. Therefore our GP as heuristic search approach can handle large sets of state variables with large numerical domains, e.g. integers., This work has been co-funded by MCIN/AEI /10.13039/501100011033 under the Maria de Maeztu Units of Excellence Programme (CEX2021-001195-M), TAILOR (H2020 #952215) and AIPlan4EU (H2020 #101016442) projects. Javier Segovia-Aguas is also supported by AGAUR SGR and the Spanish grant PID2019-108141 GB-I00. Sergio Jiménez is supported by the Spanish MINECO project PID2021-127647NB-C22. Anders Jonsson is partially supported by Spanish grant PID2019-108141 GB-I00.
Repositorio Digital de la UPF
Publicación Contribución de congreso (ConferenceOutput)

Synthesis of procedural models for deterministic transition systems

  • Segovia-Aguas, Javier
  • Ferrer Mestres, Jonathan
  • Sergio, Jiménez
Comunicació presentada a la 26th European Conference on Artificial Intelligence, celebrada del 30 de setembre al 4 d'octubre de 2023 a Kraków, Poland., This paper introduces a general approach for synthesizing procedural models of the state-transitions of a given discrete system. The approach is general in that it accepts different target languages for modeling the state-transitions of a discrete system; different model acquisition tasks with different target languages, such as the synthesis of STRIPS action models, or the update rule of a cellular automaton, fit as particular instances of our general approach. We follow an inductive approach to synthesis meaning that a set of examples of state-transitions, represented as (pre-state, action, post-state) tuples, are given as input. The goal is to synthesize a structured program that, when executed on a given pre-state, outputs its associated post-state. Our synthesis method implements a combinatorial search in the space of well-structured terminating programs that can be built using a Random-Access Machine (RAM), with a minimalist instruction set, and a finite amount of memory. The combinatorial search is guided with functions that asses the complexity of the candidate programs, as well as their fitness to the given input set of examples., This work is supported by the H2020 AIPLAN4EU project #101016442. Javier Segovia-Aguas is funded by TAILOR, AGAUR SGR and the Spanish grant PID2019-108141GB-I00. Jonathan Ferrer-Mestres is supported by the CSIRO MLAI Future Science Platform. Sergio Jiménez is funded by the Spanish MINECO project PID2021-127647NB-C22.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Learning cost action planning models with perfect precision via constraint propagation

  • Garrido, Antonio|||0000-0002-8219-6777
[EN] Data-driven AI is rapidly gaining importance. In the context of AI planning, a constraint programming formulation for learning action models in a data-driven fashion is proposed. Data comprises plan observations, which are automatically transformed into a set of planning constraints which need to be satisfied. The formulation captures the essence of the action model and unifies functionalities that are individually supported by other learning approaches, such as costs, noise/uncertainty on actions, information on intermediate state observations and mutex reasoning. Reliability is a key concern in data-driven learning, but existing approaches usually learn action models that can be imprecise, where imprecision here is an error indicator of learning something incorrect. On the contrary, the proposed approach guarantees reliability in terms of perfect precision by using constraint propagation. This means that what is learned is 100% correct (i.e., error-free), not only for the initial observations, but also for future observations. To our knowledge, this is a novelty in action model learning literature. Although perfect precision might potentially limit the amount of learned information, the exhaustive experiments over 20 planning domains show that such amount is comparable, and even better, to ARMS and FAMA, two state-of-the-art benchmarks in action model learning., This work has been partially supported by grant PID2021-127647NB-C22 funded by MCIN/AEI/10.13039/501100011033 and by "ERDF A way of making Europe", and by the Spanish MINECO project TIN2017-88476-C2-1-R.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Multimodal segmentation and labeling of lecture recordings for educational content retrieval

  • Sapena Vercher, Oscar|||0000-0003-0533-5603
  • Onaindia De La Rivaherrera, Eva|||0000-0001-6931-8293
[EN] E-learning has transformed the educational landscape, particularly in higher education, by offering flexible, scalable, and often more accessible learning environments. Lecture recordings, in particular, have become a widely used resource, offering students the ability to revisit class content at their own pace. However, the sheer volume and length of these recordings can make it difficult for learners to locate specific types of instructional content efficiently. This paper presents a hierarchical multimodal approach to segment and classify lecture recordings based on the nature of the teaching activity taking place. The proposed method integrates audio processing techniques and natural language understanding models to distinguish between various communicative functions, such as content delivery, task explanation, or organizational announcements. By leveraging both acoustic and textual cues, the system enables more effective navigation through educational videos, facilitating targeted access to relevant material. Experimental results demonstrate high overall accuracy and notable improvements over existing approaches, especially in identifying structured instructional discourse. Nonetheless, challenges persist in detecting informal or less clearly defined interactions. This work aims to enhance the usability of recorded lectures and support more personalized and efficient learning experiences., Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. This work was supported by the Spanish Ministry of Science and Innovation (MINECO) under project PID2021-127647NBC22. Funding for open access charge: CRUE-Universitat Politècnica de València.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

SmartTur+ECO: A Multimodal Conversational Recommender System Integrating Virtual Reality and Gamification for Smart Tourism

  • Carrillo-Moraga, Juan Carlos
  • Sebastiá Tarín, Laura|||0000-0001-8505-9471
  • Onaindia De La Rivaherrera, Eva|||0000-0001-6931-8293
  • Beltrán, Victoria
  • Seguí, Ignacio
  • López, Alfonso
  • Sola, Jose Carlos
[EN] This paper presents SmartTur+ECO, a system for tourist attractions that integrates conversational recommendation systems (CRS), virtual reality (VR) and gamification. It efficiently gathers trip data and elicits user preferences through interactions with a chatbot, addressing the user-cold start problem common in tourism. Using an exploit-explore approach with sentence embeddings from POI descriptions, SmartTur+ECO provides precise and personalized recommendations organized into multi-day plans, considering opening hours and locations. VR integration offers immersive pre-trip experiences, helping users make informed decisions. Gamification elements like mini-games and rewards are designed to foster user engagement and motivation. SmartTur+ECO's modular and scalable architecture ensures compatibility with various devices and allows for the easy integration of new algorithms and data sources, with the aim of supporting recommendation processes. This innovative combination sets new standards in tourism by making travel planning more efficient, entertaining and meaningful., This work is partially supported by the Spanish Research Agency project PID2025-171012NB-C21, and the Generalitat Valenciana project PROMETEO CIPROM/2023/23. This work is funded by the Valencian Innovation Agency project INNEST/2021/233. Funding for open access charge: CRUE-Universitat Politècnica de València.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

A unified constraint-based approach for plan and goal recognition from unreliable observations

  • Garrido, Antonio|||0000-0002-8219-6777
[EN] Given a sequence of observations over a plan execution, plan and goal recognition are considered as interchangeable tasks in AI planning. However, strictly speaking, the former tries to identify a plan, and the latter a set of goals, that explain the observations. Both recognition tasks are data-driven, where data comprises the plan observations, and are specially useful in proactive systems. Depending on the source of knowledge about the agents under observation, these tasks are traditionally solved by two different approaches, which require a large plan library or a planning model. In between these approaches, we propose a unified novel constraint-based approach, which distinguishes between the two tasks but is valid for both. We present a formulation, based on Partial Order Causal Link planning, that is compiled from a small plan library, to approximate a model that learns the essential causality of the original planning model. We deal with unreliable observations, which include missing and noisy observations on the real world. Modeling the observations in our formulation is straightforward. The use of the learned model allows us to address a data-driven optimization task to find the plans that most satisfy those observations (plan recognition) and the goals that are sufficiently supported by the causal relationships of the observations (goal recognition). We perform a complete evaluation of our approach in IPC domains under several indicators (accuracy, spread and ROC curves) with varying degrees of partial observability and noise on the observations. We also perform a comparison with other model-based approaches from literature., This work has been partially supported by grant PID2021- 127647NB-C22 funded by MCIN/AEI/10.13039/501100011033 and by "ERDF A way of making Europe", and by the Spanish MINECO project TIN2017-88476-C2-1-R.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Contribución de congreso (ConferenceOutput)

Learning Causality under Uncertainty for Egocentric Action Anticipation

  • Garrido, Antonio|||0000-0002-8219-6777
[EN] Egocentric action anticipation, as the task to infer future actions based on egocentric observations, is very appealing to discover patterns, conducts and, eventually, action causality. In the real world, the observations are usually uncertain, which means they are incomplete and noisy. Consequently, addressing uncertainty implies some kind of optimization. Traditional approaches develop classification tasks to calculate the actions with the highest probability to appear next (maximization problem), but they do not exploit the cause-and-effect relationships. We present here a simple two-step approach. The first step formulates and solves a constraint optimization problem that learns the action causality from a training dataset. Given an observation segment of actions over a new dataset, the second step uses such causality in a matching process to anticipate the highest scored actions. Additionally, this causality can be used to derive the planning action model. The experiments in the EPIC-KITCHENS dataset, a large-scale egocentric benchmark, show that our approach is competitive in terms of accuracy and recall., This work has been partially supported by grant PID2021-127647NB-C22 funded by MCIN/AEI/10.13039/501100011033 and by "ERDF A way of making Europe".
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Solving the tourist trip design problem with time windows and variable profit using incremental local search

  • Marzal Calatayud, Eliseo Jorge|||0000-0003-0658-1953
  • Sebastiá Tarín, Laura|||0000-0001-8505-9471
[EN] The Team Orienteering Problem with Time Windows and Variable Profits (TOPTWVP) is a variant of the Orienteering Problem where each node has a predefined time window in which the service must start (in case this node is visited), and the vehicle may spend an amount of time given by a predefined interval so that the profit collected at this node depends on the time spent. Our previous work introduced an incremental local search procedure to solve this problem. Although the results were satisfying in terms of score, this algorithm required more time to obtain solutions due to the added complexity of working with service time intervals. To address this issue, this paper introduces several improvements to our previous implementation aimed at reducing the execution time and maintaining or increasing the quality of the solutions. Specifically, we designed a new procedure for obtaining the initial solution, which is faster and generates higher -quality solutions than the previous approach. Moreover, two new moves have been developed within the local search procedure to account for variable profits. The experiments will show that these new procedures outperform our previous implementations. Finally, we define different types of trips depending on the travel -style preferences of a tourist, and we build some scenarios to show how our TOPTWVP can be adapted to obtain routes that fulfil these preferences., This work is partially supported by the Spanish Research Agency project PID2021-127647NB-C22, the Valencian Innovation Agency project INNEST/2021/233 and the EU ICT-48 2020 project TAILOR (No. 952215) .
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Automated feature extraction for planning state representation

  • Sapena Vercher, Oscar|||0000-0003-0533-5603
  • Onaindia De La Rivaherrera, Eva|||0000-0001-6931-8293
  • Marzal Calatayud, Eliseo Jorge|||0000-0003-0658-1953
[EN] Deep learning methods have recently emerged as a mechanism for generating embeddings of planning states without the need to predefine feature spaces. In this work, we advocate for an automated, cost-effective and interpretable approach to extract representative features of planning states from high-level language. We present a technique that builds up on the objects type and yields a generalization over an entire planning domain, enabling to encode numerical state and goal information of individual planning tasks. The proposed representation is then evaluated in a task for learning heuristic functions for particular domains. A comparative analysis with one of the best current sequential planner and a recent ML-based approach demonstrate the efficacy of our method in improving planner performance. Resumen Los me<acute accent>to dos de aprendizaje profundo han surgido recientemente como un mecanismo para generar embeddings de estados de planificacio<acute accent>n sin la necesidad de predefinir espacios de caracter & imath;<acute accent>sticas. En este trabajo, abogamos por un enfo que automatizado, eficiente en costes e interpretable para extraer caracter<acute accent>& imath;sticas representativas de los estados de planificacio<acute accent>n a partir de un lenguaje de alto nivel. Presentamos una te<acute accent>cnica que se basa en los tipos de objetos y permite una generalizacio<acute accent>n sobre todo un dominio de planificacio<acute accent>n, posibilitando la codificacio<acute accent>n de informacio<acute accent>n nume<acute accent>rica del estado y de los objetivos de tareas de planificacio<acute accent>n individuales. La representacio<acute accent>n propuesta se evalu<acute accent>a mediante una tarea de aprendizaje de funciones heur<acute accent>& imath;sticas para dominios espec<acute accent>& imath;ficos. Un ana<acute accent>lisis comparativo con uno de los mejores planificadores secuenciales actuales y con un enfo que reciente basado en aprendizaje automa<acute accent>tico demuestra la eficacia de nuestro me<acute accent>todo para mejorar el rendimiento de los planificadores., This work has been partially supported by the project I+D+i AEI PID2021-127647NB-C22 funded by MICIU/AEI/10.13039/501100011033 and by FEDER, UE.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Generalized planning as heuristic search: A new planning search-space that leverages pointers over objects

  • Segovia-Aguas, Javier
  • Jonsson, Anders
  • Jiménez-Celorrio, Sergio|||0000-0003-0561-4880
[EN] Planning as heuristic search is one of the most successful approaches to classical planning but unfortunately, it does not trivially extend to Generalized Planning (GP); GP aims to compute algorithmic solutions that are valid for a set of classical planning instances from a given domain, even if these instances differ in their number of objects, the initial and goal configuration of these objects and hence, in the number (and possible values) of the state variables. State -space search, as it is implemented by heuristic planners, becomes then impractical for GP. In this paper we adapt the planning as heuristic search paradigm to the generalization requirements of GP, and present the first native heuristic search approach to GP. First, the paper introduces a new pointerbased solution space for GP that is independent of the number of classical planning instances in a GP problem and the size of those instances (i.e. the number of objects, state variables and their domain sizes). Second, the paper defines an upgraded version of our GP algorithm, called Best -First Generalized Planning (BFGP), that implements a best -first search in our pointer -based solution space for GP. Lastly, the paper defines a set of evaluation and heuristic functions for BFGP that assess the structural complexity of the candidate GP solutions, as well as their fitness to a given input set of classical planning instances. The computation of these evaluation and heuristic functions does not require grounding states or actions in advance. Therefore our GP as heuristic search approach can handle large sets of state variables with large numerical domains, e.g. integers., This work has been co-funded by MCIN/AEI/10.13039/501100011033 under the Maria de Maeztu Units of Excellence Programme (CEX2021-001195-M) , TAILOR (H2020 #952215) and AIPLAN4EU (H2020 #101016442) projects. Javier Segovia-Aguas is also supported by AGAUR SGR and the Spanish grant PID2019-108141 GB-I00. Sergio Jimenez is supported by the Spanish MINECO project PID2021-127647NB-C22. Anders Jonsson is partially supported by Spanish grant PID2019-108141 GB-I00.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

A hybrid approach for expressive numeric and temporal planning with control parameters

  • Sapena Vercher, Oscar|||0000-0003-0533-5603
  • Onaindia De La Rivaherrera, Eva|||0000-0001-6931-8293
  • Marzal Calatayud, Eliseo Jorge|||0000-0003-0658-1953
[EN] Since the release of PDDL 2.1, many planners accepted the challenge of addressing more complicated domains and problems, including both temporal and numeric features for scheduling and resources management. However, very few planners can handle a significant subset of the features that the language allows for modeling. To handle a level of expressiveness sufficient to work with more realistic problems, we have designed NextFLAP (Numeric EXpressions in TFLAP). The planner follows a hybrid approach, which combines a forward partial-order planner with a numeric constraint solver/optimizer. Additionally, we include a new feature, the control parameters, which are relevant for direct modeling numerical quantities and physical properties of objects in the actions, such as ¿partially charging of a battery¿ or ¿obtaining a certain amount of resources¿. This feature overcomes a limitation of the PDDL language, which is that the parameters of the actions must be objects (finite domain values) and, in no case, numeric variables. In this paper we illustrate the advantages of using these temporal and numeric modelling features and then describe our planning system, NextFLAP, that can handle these domains. We show that NextFLAP is competitive with other state-of-the-art planners and that it generates good-quality plans., This work is supported by the Spanish AEI PID2021-127647NB-C22 project and EU ICT-48 2020 project TAILOR (No. 952215) and EU-H2020 project AIPlan4EU (101016442).
Recercat. Dipósit de la Recerca de Catalunya
Publicación Contribución de congreso (ConferenceOutput)

Synthesis of procedural models for deterministic transition systems

  • Segovia-Aguas, Javier
  • Ferrer Mestres, Jonathan
  • Sergio, Jiménez
Comunicació presentada a la 26th European Conference on Artificial Intelligence, celebrada del 30 de setembre al 4 d'octubre de 2023 a Kraków, Poland., This paper introduces a general approach for synthesizing procedural models of the state-transitions of a given discrete system. The approach is general in that it accepts different target languages for modeling the state-transitions of a discrete system; different model acquisition tasks with different target languages, such as the synthesis of STRIPS action models, or the update rule of a cellular automaton, fit as particular instances of our general approach. We follow an inductive approach to synthesis meaning that a set of examples of state-transitions, represented as (pre-state, action, post-state) tuples, are given as input. The goal is to synthesize a structured program that, when executed on a given pre-state, outputs its associated post-state. Our synthesis method implements a combinatorial search in the space of well-structured terminating programs that can be built using a Random-Access Machine (RAM), with a minimalist instruction set, and a finite amount of memory. The combinatorial search is guided with functions that asses the complexity of the candidate programs, as well as their fitness to the given input set of examples., This work is supported by the H2020 AIPLAN4EU project #101016442. Javier Segovia-Aguas is funded by TAILOR, AGAUR SGR and the Spanish grant PID2019-108141GB-I00. Jonathan Ferrer-Mestres is supported by the CSIRO MLAI Future Science Platform. Sergio Jiménez is funded by the Spanish MINECO project PID2021-127647NB-C22.
Recercat. Dipósit de la Recerca de Catalunya
Publicación Artículo científico (JournalArticle)

Generalized planning as heuristic search: a new planning search-space that leverages pointers over objects

  • Segovia-Aguas, Javier
  • Jiménez, Sergio
  • Jonsson, Anders, 1973-
Planning as heuristic search is one of the most successful approaches to classical planning but unfortunately, it does not trivially extend to Generalized Planning (GP); GP aims to compute algorithmic solutions that are valid for a set of classical planning instances from a given domain, even if these instances differ in their number of objects, the initial and goal configuration of these objects and hence, in the number (and possible values) of the state variables. State-space search, as it is implemented by heuristic planners, becomes then impractical for GP. In this paper we adapt the planning as heuristic search paradigm to the generalization requirements of GP, and present the first native heuristic search approach to GP. First, the paper introduces a new pointer-based solution space for GP that is independent of the number of classical planning instances in a GP problem and the size of those instances (i.e. the number of objects, state variables and their domain sizes). Second, the paper defines an upgraded version of our GP algorithm, called Best-First Generalized Planning (BFGP), that implements a best-first search in our pointer-based solution space for GP. Lastly, the paper defines a set of evaluation and heuristic functions for BFGP that assess the structural complexity of the candidate GP solutions, as well as their fitness to a given input set of classical planning instances. The computation of these evaluation and heuristic functions does not require grounding states or actions in advance. Therefore our GP as heuristic search approach can handle large sets of state variables with large numerical domains, e.g. integers., This work has been co-funded by MCIN/AEI /10.13039/501100011033 under the Maria de Maeztu Units of Excellence Programme (CEX2021-001195-M), TAILOR (H2020 #952215) and AIPlan4EU (H2020 #101016442) projects. Javier Segovia-Aguas is also supported by AGAUR SGR and the Spanish grant PID2019-108141 GB-I00. Sergio Jiménez is supported by the Spanish MINECO project PID2021-127647NB-C22. Anders Jonsson is partially supported by Spanish grant PID2019-108141 GB-I00.