CA-INP contributes to different WPs in the project. For WP2 and 4, we successfully managed to handle textureless objects using a 3D camera. We introduced a new tracking and a new servoing pipeline which can be implemented in practical industrial cases where there is no specific texture like the Michelin use case. Furthermore, we developed 3D shape reconstruction using Digit tactile sensor and tested this method in the doll assembly task. Also, we designed dual quaternion-based dynamic movement primitives (DMP) to learn bimanual tasks. We compared its performance with traditional decoupled DMP with a tire tread deformation task. We also added shape feedback to improve shape servoing.

Finally and for WP5 and 6, we integrated these techniques on Sigma mockups and currently working with Michelin to integrate all Sigma’s technical contributions to their robotic cell.