Search-Based Task Planning with Learned Skill Effect Models for Lifelong Robotic Manipulation
Jacky Liang*, Mohit Sharma*, Alex LaGrassa, Shivam Vats, Saumya Saxena, Oliver Kroemer
* Equal Contribution
Task A
Pick-Place + Tray Slide + Tray-Sweep + Bin-Tilt
Task B
Pick-Place + Tray Slide + Tray-Sweep (Left: Scenario 1, Right: Scenario: 2)
Task D
Pick-Place+Tray-Slide + Tray-Sweep + Bin-Tilt (Left: Scenario 1, Right: Scenario 2)
Real World Results (Task B)
Pick-Place
Pick-Place + Tray-Slide
Pick-Place + Tray-Slide + Tray-Sweep
Size Generalization for Task-B
Object size: 0.02 with Pick-Place
Object size: 0.02,
Pick-Place+Tray-Slide
Object size: 0.03, with Pick-Place
Object size: 0.03
Pick-Place + Tray-slide
Color Generalization for Task-B
One block with red color,
Pick-Place + Tray-Slide skills
Two blocks with red color,
Pick-Place skill only
Two blocks with red color,
Pick-Place + Tray-Slide skills