3D grasping solutions through MWS models

In order to solve a great number of robotics problems in 3D environments, raw information yielded by visual sensors can be discretized imposing a 3D model with a specific mesh topology. The goal of this work is to use the solid representation model MWS (modeling wave set) for solving grasping problems on 3D free shape objects. This implies to reinforce the connection between the fields of modeling and 3D-grasping. Through the concept of direction kernels, integrated in the model MWS, we define the interaction regions and the criteria for evaluating an optimal grasp. The set of direction kernels is a global feature of the object which represents their main normal directions and their relative weight in the mesh model. The key property of such a model to be used in robot interaction applications consists of the direction kernels invariance to affine transformations. This strategy has been experimented with real objects sensed with a 3D laser-scanner. The system successfully calculates the optimal grasp points, ordered according to the quality function, for two-finger grippers maintaining the requirements of form closure and safety and stability of the grasp. The obtained results may lead to the successful application of this strategy in systems with a greater number of grasp contacts