Control and design of robots with tasks and constraints in mind - Archive ouverte HAL Accéder directement au contenu
Hdr Année : 2016

Control and design of robots with tasks and constraints in mind

Résumé

The work presented in this dissertation is mostly concerned with the problem of controlling robots. While a large part of the scientific literature in Robotics is dedicated to this problem, there still exists a gap between what has been proposed using advanced control techniques and the majority of existing applications. Indeed, in simulation or in lab conditions, ad hoc environment and situations can be generated in order to simplify the control problem and only address one of its sub-parts. This can be necessary in the preliminary stages of research, when trying to address a challenging problem. However, the risk is to provide solutions which are intrinsically incompatible with real life constraints. These constraints often constitute strong non-linearities which have to be accounted for to maintain the robot and its surrounding environment in proper working conditions. Nevertheless, these limits have often been dealt with as exceptions or secondary objectives. If these approaches make sense in cases where all motions can be pre-planned, they do not apply to situations where motions have to be generated in reaction to the environment. In that sense, there is still a need for advanced control techniques that intrinsically account for real world constraints. These constraints both lie in the actuation and configuration spaces of the robot but also in the tasks spaces, i.e. in spaces where the tasks assigned to the robot can be expressed in a straightforward way by the programmer or by the user through interactive programming. Thus, robotics control paradigms have to provide the means to naturally express and optimally perform concurrent tasks while natively accounting for constraints in a computationally efficient way. In an attempt to tackle this general problem, the first research direction presented in this dissertation describes some work performed in constraints compatible, multi-tasks robot control. The proposed contribution in that domain leads to the formulation of the robotics control problem as a constrained optimisation one. This formulation provides the benefit of computing locally optimal solutions intrinsically compatible with the constraints. This is particularly suitable for applications in constrained and dynamic environments. While the first chapter advocates for the formulation of the reactive control problem as an optimisation one, the limits of reactive approaches are quite tangible when dealing with the constraints compatibility problem. Global optimality has to be tackled despite its complexity. This complexity reaches its climax with humanoid robots and the second contribution presented in this dissertation is centred on a review of optimisation-based control approaches for balanced humanoid behaviours. Control is the central focus of the presented work. Nonetheless, designing the right robot for a given application is also a very interesting and important research topic if one envisions less traditional use of robots with respect to standard manufacturing applications. Thus, the final contribution presented in this document is related to the performance based design and evaluation of robots with collaborative applications in mind.
Fichier principal
Vignette du fichier
memoire_HDR_Vincent_Padois_vf.pdf (22.57 Mo) Télécharger le fichier

Dates et versions

tel-01398868 , version 1 (17-11-2016)

Identifiants

  • HAL Id : tel-01398868 , version 1

Citer

Vincent Padois. Control and design of robots with tasks and constraints in mind. Robotics [cs.RO]. Université Pierre et Marie Curie (Paris 6), 2016. ⟨tel-01398868⟩
315 Consultations
357 Téléchargements

Partager

Gmail Facebook X LinkedIn More