base::OptimizationObjective Class Reference

Abstract definition of optimization objectives. More...

#include <ompl/base/OptimizationObjective.h>

Inheritance diagram for base::OptimizationObjective:

List of all members.

Public Member Functions

 OptimizationObjective (const SpaceInformationPtr &si)
 Constructor. The objective must always know the space information it is part of. The cost threshold for objective satisfaction defaults to 0.0.
const std::string & getDescription () const
 Get the description of this optimization objective.
virtual bool isSatisfied (Cost c) const
 Verify that our objective is satisfied already and we can stop planning.
Cost getCostThreshold () const
 Returns the cost threshold currently being checked for objective satisfaction.
void setCostThreshold (Cost c)
 Set the cost threshold for objective satisfaction. When a path is found with a cost better than the cost threshold, the objective is considered satisfied.
virtual bool isCostBetterThan (Cost c1, Cost c2) const
 Check whether the the cost c1 is considered better than the cost c2. By default, this returns true only if c1 is less by at least some threshold amount, for numerical robustness.
virtual Cost stateCost (const State *s) const =0
 Evaluate a cost map defined on the state space at a state s.
virtual Cost motionCost (const State *s1, const State *s2) const =0
 Get the cost that corresponds to the motion segment between s1 and s2.
virtual Cost combineCosts (Cost c1, Cost c2) const
 Get the cost that corresponds to combining the costs c1 and c2. Default implementation defines this combination as an addition.
virtual Cost identityCost () const
 Get the identity cost value. The identity cost value is the cost c_i such that, for all costs c, combineCosts(c, c_i) = combineCosts(c_i, c) = c. In other words, combining a cost with the identity cost does not change the original cost. By default, a cost with the value 0.0 is returned. It's very important to override this with the proper identity value for your optimization objectives, or else optimal planners may not work.
virtual Cost infiniteCost () const
 Get a cost which is greater than all other costs in this OptimizationObjective; required for use in Dijkstra/Astar. Defaults to returning the double value inf.
virtual Cost initialCost (const State *s) const
 Returns a cost value corresponding to starting at a state s. No optimal planners currently support this method. Defaults to returning the objective's identity cost.
virtual Cost terminalCost (const State *s) const
 Returns a cost value corresponding to a path ending at a state s. No optimal planners currently support this method. Defaults to returning the objective's identity cost.
virtual bool isSymmetric () const
 Check if this objective has a symmetric cost metric, i.e. motionCost(s1, s2) = motionCost(s2, s1). Default implementation returns whether the underlying state space has symmetric interpolation.
virtual Cost averageStateCost (unsigned int numStates) const
 Compute the average state cost of this objective by taking a sample of numStates states.
void setCostToGoHeuristic (const CostToGoHeuristic &costToGo)
 Set the cost-to-go heuristic function for this objective. The cost-to-go heuristic is a function which returns an admissible estimate of the optimal path cost from a given state to a goal, where "admissible" means that the estimated cost is always less than the true optimal cost.
Cost costToGo (const State *state, const Goal *goal) const
 Uses a cost-to-go heuristic to calculate an admissible estimate of the optimal cost from a given state to a given goal. If no cost-to-go heuristic has been specified with setCostToGoHeuristic(), this function just returns the identity cost, which is sure to be an admissible heuristic if there are no negative costs.
virtual Cost motionCostHeuristic (const State *s1, const State *s2) const
 Defines an admissible estimate on the optimal cost on the motion between states s1 and s2. An admissible estimate always undervalues the true optimal cost of the motion. Used by some planners to speed up planning. The default implementation of this method returns this objective's identity cost, which is sure to be an admissible heuristic if there are no negative costs.
const SpaceInformationPtr & getSpaceInformation () const
 Returns this objective's SpaceInformation. Needed for operators in MultiOptimizationObjective.

Protected Attributes

SpaceInformationPtr si_
 The space information for this objective.
std::string description_
 The description of this optimization objective.
Cost threshold_
 The cost threshold used for checking whether this objective has been satisfied during planning.
CostToGoHeuristic costToGoFn_
 The function used for returning admissible estimates on the optimal cost of the path between a given state and goal.

Detailed Description

Abstract definition of optimization objectives.

Note:
This implementation has greatly benefited from discussions with Kris Hauser

Definition at line 66 of file OptimizationObjective.h.


The documentation for this class was generated from the following files:
 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Defines