« Simulation

This page contains simulation videos related to the NAMO papers. Please contact Mike Stilman for live robot movies, additional simulations and further information about this project.

« ICRA (2007) » Manipulation Planning

This work extends our domain to articulated robots in three dimensional environments. The robot is provided with kinematic and geometric models of itself and scene objects. Our algorithm chooses stable placements for objects and uses multi-goal RRTs to sample high dimensional paths for navigation and manipulation. Based on our proposed method at WAFR'06, we construct models of space that is necessary for future motion samples. The algorithm uses these models to identify objects that must be moved and constrain their placements.

« WAFR (2006) » Planning with Artificial Constraints & Execution

These two videos illustrate the algorithm for planning with Artificial Constraints. Once a future displacement path is decided, the swept volume for this path is used as a constraint on the placements of colliding objects. The planner then constructs corresponding displacements. In the first movie, notice that when the planner cannot displace the couch it backtracks and selects a different ordering of objects. The second video illustrates the final execution.

« Humanoids (2004) »

A simulation of our vision for Navigation Among Movable Obstacles. The robot is required to reach a door at the back of the scene. It must decide which objects must be moved, where to move them and construct plans for navigation and manipulation. The planner operates in a 2D projection of the world. While simulated dynamics for walking are fully controlled, simulated manipulation is simplified and does not take into account grasping or balance.

Two sample NAMO problems solved with our first algorithm that reasons about configuration space connectivity. The fist problem contains nearly 100 movable objects. By heuristically identifying keyholes in the plan space we focus search towards manipulation that makes progress in the overall task.

Copyright 2004-2007 Mike Stilman.