Contribution a la modélisation dynamique d’un robot flexible bionique

Abstract

This thesis deals with the design, geometric, kinematic and dynamic modeling of bionic flexible robots. In particular, we are mainly interested in cable-driven flexible robots and a bionic robot named ‘‘Compact Bionic Handling Assistant (CBHA)’’. Initially, we have proposed two kind of design of flexible robots, including a planar robot and a space robot, both powered by cables. Based on this study, we have undertaken the construction of a prototype with a single flexible section. This design and realization have also been used to estimate the inertial parameters that will be used in the dynamic models of these prototypes. First, we have developed the mathematical formulations describing the structure of the flexible robots which represents a general case of multi-section flexible robots. Next, we have developed an approach for solving the problem of inverse geometric modeling of multi-section flexible robots, assuming that each flexible section bends as a circular arc while keeping the principal axis of the structure inextensible. The problem was formulated mathematically in terms of optimization of a quadratic cost function under some constraints of equality (lengths conservation). The problem has been solved by developing an optimization algorithm including metaheuristic methods, namely the Particle Swarm Optimization (PSO) and the Genetic Algorithm (GA). This approach was validated by simulation and two experimental test benches performed on the CBHA robot. Then, we have also proposed dynamic models, direct and inverse, for the two cable-driven flexible robots, planar and spatial, using the Lagrange method. In order to simplify the calculations and avoid some numerical singularities, the dynamic models were approximated by expansion of Taylor's series. These models have been validated by simulations under the MATLAB environment. In addition, the inverse dynamic model was validated by real measurements obtained from the robot UR5. Finally, a proportional-integral-derivative controller (PID) has been proposed to track trajectories using the point-to-point method.