Research ArticleARTIFICIAL INTELLIGENCE

Vision-based grasp learning of an anthropomorphic hand-arm system in a synergy-based control framework

See allHide authors and affiliations

Science Robotics  30 Jan 2019:
Vol. 4, Issue 26, eaao4900
DOI: 10.1126/scirobotics.aao4900
  • Fig. 1 Schematic overview of the proposed algorithm.
  • Fig. 2 Example of point cloud processing.

    (A) Original point cloud. (B) Processed point cloud. The main plane in the scene has been removed, and the segmentation of the remainder of the cloud has been executed.

  • Fig. 3 Experiment on a tennis ball.

    (A) Grasp success table. (B) Force-closure cost function. (C) Synergy coefficients. (D) Arm scores.

  • Fig. 4 Experiment on a plastic strawberry.

    (A) Grasp success table. (B) Force-closure cost function. (C) Synergy coefficients. (D) Arm scores.

  • Fig. 5 Experiment on a plastic bottle.

    (A) Grasp success table. (B) Force-closure cost function. (C) Synergy coefficients. (D) Arm scores.

  • Fig. 6 Successful grasps of the tennis ball, strawberry, and plastic bottle at the end of the learning process.
  • Table 1 Comparison of grasps obtained with and without force-closure cost function.
    Force-closure cost value
    GraspCost not usedCost used
    Bipodal1.9 × 1078.5 × 105
    Tripodal2.5 × 1073.1 × 106
    Sphere five finger2.7 × 1084.6 × 107
    Cylinder five finger1.9 × 1081.4 × 107

Stay Connected to Science Robotics

Navigate This Article