Design And Grasp Planning Of Synergy-based Anthropomorphic Robotic Hands

Postural synergy originally refers to a coordinated contraction pattern of a group of muscles and the concept of synergy has triggered many research activities.In researches concerning hands and grasping,the postural synergies provide new understanding to hand motion atlas that used to primarily consist of discrete grasp taxonomy.In addition,the synergies also enabled new approaches for grasp planning and in-hand manipulation.This dissertation presents investigations on the design and grasp planning of anthropomorphic robotic hands,including a mechanical postural synergy implementation,as well as a grasp planning framework composed of a gradient-based iterative planning formulation,a stability-based predictive reach-to-grasp algorithm,and a stereo-vision-based surface reconstruction method.While implementing postural synergies mechanically for an anthropomorphic robotic hand,a CCM(Composed Continuum Mechanism)was proposed and used as a transmission unit to generate an arbitrary number of translating outputs that are linearly combined from two independent linear translational inputs.In this hand,three actuators actuate the CCM in its coordinated motion mode to drive the eleven hand joints according to two synergy inputs to form synergy-based hand poses in a pre-grasp phase.Then,the CCM closes the fingers in its synchronized motion mode to form a stable grasp.Joint-level compliance was introduced based on a statics analysis to help achieve the stable grasps and pinches of many daily-life objects.The compliance calculation,postural synergy synthesis,system descriptions and experimental characterizations of the hand with Compliant Mechanical Postural Synergy(the Co MPS hand)are detailed.In order to plan reach-to-grasp motions for a synergy-based anthropomorphic robotic hand,a unified kinematics framework is proposed based a gradient-based iterative formulation,predicting the grasp quality and adjusting the postural synergies,the position and the orientation of the hand during the approaching phase to achieve a stable grasp via reduced trials.The pseudo-distance function is firstly extended to 3D Euclidian space to guide the hand to approach the object,and then used in 6D wrench space to predict the grasp quality.Stable power grasps and precision pinches can be consistently and intentionally planned with different contact conditions specified in the formulation.Several numerical simulation case studies are presented to demonstrate the effectiveness of the proposed algorithm.Furthermore,a stereo vision system is used for imaging acquisition for reconstructing the to-be-grasped object surfaces from point clouds using an incremental Structure from Motion(SFM)algorithm.This algorithm necessarily enables the proposed grasp planning algorithm to be applied in a practical setting.The surface reconstruction algorithm consists of three aspects: stereo disparity map calculation,ego-motion estimation based on feature matching,and 3d reconstruction after optimizing the re-projection errors.Numerical simulation experiments are presented to verify the effectiveness.With the design of the anthropomorphic robotic hand based on the compliant mechanical postural synergies and the reach-to-grasp planning algorithm based on the stereo vision system,the postural synergies have been shown efficiently in the understanding of the hand motion atlas and the grasp planning for achieving various realistic reach-to-grasp tasks.