Generic Instructions for Setting Up a Planning Context

Instantiating a state space

The first step for setting up a planning problem is selecting the space we are planning in. Simply create an instance of a class that inherits from ompl::base::StateSpace, or select one from the available implementations. To change the definition of the employed sampler or distance function, it is possible to provide a further specialization of such a state space. See [Implementing State Spaces]{implementingStateSpaces.html}.

base::StateSpacePtr space(new base::SE2StateSpace());
// set bounds ....

Instantiating a control space

If planning with differential constraints, a representation of the space of controls is needed as well. Simply create an instance of a class that inherits from ompl::control::ControlSpace. Usually, ompl::control::RealVectorControlSpace is sufficient. An implementation needs to be specified for the ompl::control:ControlSpace::propagate() either by inheriting from the control space class or by calling ompl::control::ControlSpace::setPropagationFunction().

base::StateSpacePtr space(new base::SE2StateSpace());
// set bounds for state space
control::ControlSpacePtr cspace(new control::RealVectorControlSpace(space));
// set bounds for cspace

Instantiating a space information class

Creating an actual instance of a space information class (ompl::base::SpaceInformation) is trivial as the constructor requires only a state space to be specified (ompl::base::StateSpace). When planning with controls (ompl::control::SpaceInformation), this constructor requires a control space (ompl::control::ControlSpace) as well. The space information class also needs to be configured before use:

  • ompl::base::StateValidityChecker is an abstract class that provides functionality for determining whether a state is valid or not. This class is assumed to be thread safe. The user should provide an implementation of this class and supply it to the space information instance by calling ompl::base::SpaceInformation::setStateValidityChecker(). Alternatively, the user can pass a function of the type ompl::base::StateValidityCheckerFn to ompl::base::SpaceInformation::setStateValidityChecker() instead. By default, all states are considered valid, if this parameter is not set.
  • if using ompl::base::DiscreteMotionValidator for validating motions (this is the default), a call needs to be made to ompl::base::SpaceInformation::setStateValidityCheckingResolution() in order to specify the maximum distance between states to be checked for validity along a path segment. This distance is specified as a percentage of a space's maximum extent. If this call is not made, the resolution is assumed to be 1%. This value may be too low, in which case planning will be slower, or it may be too high, in which case it is possible to have collisions in solution plans.
    // define this class:
    class myStateValidityCheckerClass : public base::StateValidityChecker
    myStateValidityCheckerClass(const base::SpaceInformationPtr &si) :
    virtual bool isValid(const base::State *state) const
    return ...;
    // or this function:
    bool myStateValidityCheckerFunction(const base::State *state)
    return ...;
    base::SpaceInformationPtr si(space);
    // either this call:
    // or this call:
    si->setStateValidityCheckingResolution(0.03); // 3%
    Once the class is instantiated and parameters have been set, the ompl::base::SpaceInformation::setup() function needs to be called and the instance is ready for use.

Instantiating a problem definition

Instantiating a planner

In order to use a motion planner (ompl::geometric::XXX, from ompl/geometric/planners or ompl::control::XXX, from ompl/control/planners), an instance of ompl::base::SpaceInformation (ompl::control::SpaceInformation, respectively) must be available. This instance is supplied to the planner's constructor. After creation, a call to the planner's ompl::base::Planner::setup() method must be made and the planner instance is ready for use. The problem to be solved is set with ompl::base::Planner::setProblemDefinition(). The ompl::base::Planner::solve() method can be called repeatedly with different allowed time durations until a solution is found. The planning process continues with the available data structures when sequential calls to ompl::base::Planner::solve() are made. A call to ompl::base::Planner::clear() will restore a planner to its state before any calls to the ompl::base::Planner::solve() method were made.

using namespace ompl;
base::SpaceInformationPtr si(...);
// set start states & goal region for the problem definition
base::PlannerPtr planner(new geometric::SBL(si));
if (pdef->getGoal()->getSolutionPath())
// do something with the solution

Using SimpleSetup

The instantiation of all the above classes is facilitated by the ompl::geometric::SimpleSetup class (ompl::control::SimpleSetup, respectively). Please see the documentation of these classes for more information. Making use of their functionality is recommended, if appropriate.

Definition of a problem to be solved. This includes the start state(s) for the system and a goal spec...
Single-Query Bi-Directional Probabilistic Roadmap Planner with Lazy Collision Checking.
Definition: SBL.h:148
Main namespace. Contains everything in this library.
Definition: AppBase.h:21