RH-TAMP

Technical Description

An overview of the Receding Horizon TAMP method. Given a target task, a geometric reasoning module iteratively evaluates the feasibility of a sequence of h+1 actions (the actions horizon) considered over the full set of n actions provided by the task planner. Hence, the motion planner guides the robot toward the first available configuration q_j, and the process restarts by moving forward of one step the actions window of size h+1. If the plan is not feasible, the action fails, or the state of the system changes unexpectedly, the task planner should be executed again.

Experiments

Clutter Table Problem

The Clutter Table task and motion planning problem. The robot hasto pick up the red cylinder placed between 25 other ones. The robot can grasp cylindersonly from their side.

Non-Monotonic Problem

The Non-Monotonic task and motion planning problem. The robot has to move the red and blue cylinders from their initial position to the positions corresponding to their color. The robot can grasp cylinders only from their side.

Sort Clutter Problem

The Sort Clutter task and motion planning problem. The mobile robot has to transport the cylinders of different colors from the initial position to the right table, sorting them by color.

Code

A software implementation of this project can be found on our GitHub repository. The implementation is based on ROS and our modification to support RH-TAMP is licensed for non-commercial use (license summary).

Technical Description

Experiments

Clutter Table Problem

Non-Monotonic Problem

Sort Clutter Problem

Code

People

Nicola Castaman

Enrico Pagello

Emanuele Menegatti

Alberto Pretto