Skip to main content

Module 07: Manipulation

Introduction​

Manipulationβ€”the ability to physically interact with and transform objectsβ€”is essential for humanoid utility. This module covers grasping, force control, and dexterous manipulation.

Section 1: Grasping​

1.1 Grasp Analysis​

Form Closure: A grasp configuration where the object cannot move regardless of applied forces (geometric constraint).

Force Closure: A grasp where any external wrench can be resisted by contact forces within friction cones.

1.2 Grasp Planning​

def plan_grasp(object_mesh, gripper_model):
    # Sample candidate grasps
    candidates = sample_antipodal_grasps(object_mesh)

    # Score by quality metrics
    scored = []
    for grasp in candidates:
        quality = compute_grasp_quality(grasp, object_mesh)
        scored.append((grasp, quality))

    # Return best feasible grasp
    return max(scored, key=lambda x: x[1])

Section 2: Force Control​

2.1 Hybrid Position/Force Control​

Partitioning task space into position and force controlled directions:

Ο„=JT(SpKpex+SfKfef)\boldsymbol{\tau} = \mathbf{J}^T(\mathbf{S}_p\mathbf{K}_p\mathbf{e}_x + \mathbf{S}_f\mathbf{K}_f\mathbf{e}_f)

2.2 Contact Stability​

Maintaining stable contact while applying force.

Force control requires accurate force sensing. Sensor noise and dynamics can cause instability at high gains.

Section 3: Dexterous Manipulation​

3.1 Multi-Finger Coordination​

Humanoid hands with 10+ DOF enable:

  • In-hand manipulation
  • Tool use
  • Precise object positioning

3.2 Tactile Sensing​

High-resolution tactile sensors provide:

  • Contact location
  • Contact force distribution
  • Slip detection

Section 4: Learning for Manipulation​

4.1 Imitation Learning​

Learning from human demonstrations.

4.2 Reinforcement Learning​

Learning manipulation skills through trial and error.

Summary​

Key takeaways:

  1. Grasp quality determines manipulation reliability
  2. Force control enables contact-rich tasks
  3. Dexterous hands enable human-like manipulation
  4. Learning accelerates skill acquisition

Key Concepts​

  • Force Closure: Ability to resist any external wrench
  • Hybrid Control: Simultaneous position and force control
  • Dexterity: Fine manipulation with multi-finger hands
  • Tactile Sensing: Measuring contact forces and location

Further Reading​

  1. Mason, M.T. (2001). "Mechanics of Robotic Manipulation"
  2. Prattichizzo, D. & Trinkle, J.C. (2016). "Grasping" in Handbook of Robotics