hpp-manipulation Documentation

Introduction

This package implements a solver for manipulation planning problems. A manipulation planning problem is defined by:

• Device: a collection of several Device with grippers and handles,
• Graph: A graph of constraints defining the rules of a manipulation problem.

To build a Device, have a look at the package hpp-manipulation-urdf. ManipulationPlanner implements a RRT-based algorithm to solve manipulation planning problems.

Constraint graph

The graph of constraint, also referred to as constraint graph, represents the rules of a manipulation problem. The component of the graph are:

• Node represents a state of the Robot with constraints,
• Edge represents a transition between two Node with parametric constraints.

Node contains a set of Constraint that a configuration of the Robot should satisfy to be in the represented state. To ensure that a configuration is in only one state, the Node are ordered in a NodeSelector. The method NodeSelector::getNode(ConfigurationIn_t) const returns a pointer to the first Node for which Node::contains(ConfigurationIn_t) const returns true. For optimization only, another set of Constraint is used for StraightPath lying in this Node.

Edge has methods Edge::isInNodeFrom, to tell if a corresponding path lyes in Edge::from() or Edge::to(), and Edge::node(), to retrive this Node. Edge also contains two sets of Constraint:

• Edge::configConstraint() returns a ConstraintSet used to generate a configuration lying in Edge::to() and respecting the ConfigProjector::rightHandSide,
• Edge::pathConstraint() returns a ConstraintSet to be inserted in Path represented by this Edge.

Note:

For implementation details, see graph::Graph. For more information about parametric and non-parametric constraints, see DifferentiableFunction and ConfigProjector

Manipulation planner

ManipulationPlanner class implements an algorithm based on RRT. See this master thesis for details about the algorithm.