EECS 106B Project3 Grasping

Contribuator: Wenzhe Tong, Anxing Xiao
Date: Apr 8th, 2020

Tasks

• Write metrics to score a grasp

  • Resistance of gravity (Done)
  • Ferrari Canny (Done)
  • Robust Force Closure (Done)

• Write our own planner

  • Sampling
  • Filtering
  • Planning

Starter Code

• grasping.py is what you need to modify
• utils.py is utilities functions you can refer and modify
• grasping video provided here: Pawn, Nozzle

How to run

• using conda create a python3.6 environment
• cvxpy, scipy, vtkplotter, trimeshpackage needed (see pdf for more info)
• Uncomment 2 lines in function main() to visualize grasp

To be mentioned

• 3rd grasping result of Nozzle is [1 1 1 1 1] instead of [0 0 0 0 0]

Result

• Grading Metric:

  • Gravity Resistance
  • Ferrari-Canny
  • Robust Force Closure
    We sampling in ${\mathbb{R}}^{\mathbb{3}}$$\mathbb{R^3}$ by normal distribuation around the contact point and do the Monte-Carlo estimation aobut the force closure probability.

Sampling in R3	Project to mesh surface

• Planning

  • Sampling
    We first sampling from mesh surface by the function randomly_sample_from_mesh and pick 2 points, feed it to the filter by distance constraints and so on.
  • Planning
    We recover the pose by really naive algorithm, choose the best sampling points and calculate the pose then visualize it.

Nozzle	Pawn

• Given Grasp:

Nozzle	Pawn