Retour au site du LAAS-CNRS

Abstract

This article presents a new open-source C++ implementation to solve the SLAM problem, which is focused on genericity, versatility and high execution speed. It is based on an original object oriented architecture, that allows the combination of numerous sensors and landmark types, and the integration of various approaches proposed in the literature. The system capacities are illustrated by the presentation of an inertial/vision SLAM approach, for which several improvements over existing methods have been introduced, and that copes with very high dynamic motions. Results with a hand-held camera are presented.

Abstract

This article presents a new open-source C++ implementation to solve the SLAM problem, which is focused on genericity, versatility and high execution speed. It is based on an original object oriented architecture, that allows the combination of numerous sensors and landmark types, and the integration of various approaches proposed in the literature. The system capacities are illustrated by the presentation of an inertial/vision SLAM approach, for which several improvements over existing methods have been introduced, and that copes with very high dynamic motions. Results with a hand-held camera are presented.

Abstract

While key for human-robot interaction, natural language interpretation is a notoriously difficult task, especially because the interaction context is at the same time essential for dialogue understanding, difficult to build for machines, and depends on each speaker point of view. However, robots as embodied artifacts, can perceive their environment and interactors, and hence compute symbolic models from various perspectives. This allows in turn to build symbolic contexts for dialogues. In this paper, we introduce DIALOGS, a component for natural language interpretation that relies on these structured symbolic models of the world to ground verbal interaction.

Résumé

Le contexte de la robotique de service est caractérisé par la présence de l'homme dans l'espace de travail du robot. Les mouvements de ces robots ne doivent perturber ni la sécurité de l'homme ni son confort. D'un point de vu planification de mouvement, le planificateur doit d'une part éviter de heurter l'homme ou l'environnement et d'autre part adapter les limites cinématiques du robot en fonction de la proximité de l'homme. A chaque niveau du système (planification et exécution/contrôle), le robot doit garantir la sécurité et le confort de l'homme. Nous proposons une approche de la planification et du contrôle de mouvement basée sur des trajectoires polynomiales. Dans une première partie, nous présentons un générateur de trajectoires qui limite la vitesse, l'accélération et le jerk. Il génère des trajectoires composées de suites de segments de courbes cubiques. Le cas mono-dimensionnel est d'abord présenté puis étendu au cas multi-dimensionnel. Dans une deuxième partie, nous proposons d'approximer les trajectoires par des suites de triplets de segments de courbes cubiques. Cette méthode permet de calculer des trajectoires respectant une erreur maximale donnée. Ces générateurs de trajectoire sont intégrés au planificateur de chemin et produisent des trajectoires directement exécutables. Une application originale de l'approximation permet d'approximer une trajectoire définie dans l'espace cartésien par une trajectoire définie dans l'espace articulaire. Cette approche simplifie la structure du contrôleur du robot. La présence de l'homme dans l'espace de travail du robot nécessite une adaptation des trajectoires pendant l'exécution. Nous proposons une méthode pour adapter la loi de mouvement de la trajectoire multidimensionnelle pendant l'exécution. Ces travaux, menés dans le cadre du projet européen DEXMART et du projet ANR ASSIST, ont été intégrés et validés sur les plateformes Jido et PR2 du LAAS-CNRS.

Abstract

The context of service robotics is characterized by the presence of humans in the vicinity of the robot. The movements of these robots should not disturb the safety of humans or their comfort. From the motion planning point of view, the planner must both avoid hitting humans or colliding with the environment and also adapt the robot's kinematic limits depending on the proximity of humans. At each level of the system (Planning and execution / Control), the robot must ensure the safety and the comfort of humans. We propose an approach of motion planning and motion control based on polynomial trajectories. In the first part, we present a trajectory generator which limits the speed, the acceleration and the jerk (derivative of the acceleration). The motion planner generates trajectories consisting of series of segments of cubic polynomial curves. The mono-dimensional case is first introduced and then extended to the multi-dimensional one. In the second part, we propose to approximate the trajectories by sequences of triplets of segments of cubic curves. This method allows to find trajectories that respect a given maximum error. These trajectory generators are integrated into the path planner and produce directly executable motion. An original application of the trajectory approximation is the approximation of a trajectory defined in Cartesian space by a trajectory defined in the joint space. This approach simplifies the structure of the robot controller. The presence of humans in the workspace of the robot requires also an adaptation of the trajectories during the execution. We propose a method to adapt the motion law of the multidimensional path at runtime. This work, conducted as part of the European project DEXMART and the ANR project ASSIST, has been integrated and validated on the Jido and PR2 platforms of LAAS-CNRS.

Mots-Clés / Keywords

Robotique; Planification de trajectoires; Interaction homme-robot; Manipulation; Robotics; Trajectory planning; Human-robot interaction;

Abstract

Interacting with a robot in a shared space requires not only a shared model (i.e., environment entities must be described with the same symbols), but also a model of mutual knowledge: both the robot and the human need to figure out what the other knows, sees or can do in the environment to appropriately behave. We present here our efforts to let the robot grounds verbal interaction with a human in a physical, situated environment. We propose a knowledge-oriented architecture, where perceptions from different point of views (from the robot itself, from the human, etc.) are turned into symbolic facts that are stored in different cognitive models and reused in a newly designed module for dialogue grounding.