ompl::control::Syclop Class Reference

Synergistic Combination of Layers of Planning. More...

#include <ompl/control/planners/syclop/Syclop.h>

Inheritance diagram for ompl::control::Syclop:

List of all members.

Classes

class  Adjacency
 Representation of an adjacency (a directed edge) between two regions in the Decomposition assigned to Syclop. More...
class  CoverageGrid
 Syclop uses a CoverageGrid to estimate coverage in its assigned Decomposition. The CoverageGrid should have finer resolution than the Decomposition.
struct  Defaults
 Contains default values for Syclop parameters. More...
class  Motion
 Representation of a motion. More...
class  Region
 Representation of a region in the Decomposition assigned to Syclop. More...

Public Types

typedef boost::function
< double(int, int)> 
EdgeCostFactorFn
 Each edge weight between two adjacent regions in the Decomposition is defined as a product of edge cost factors. By default, given adjacent regions $r$ and $s$, Syclop uses the sole edge cost factor

\[ \frac{1 + \mbox{sel}^2(r,s)}{1 + \mbox{conn}^2(r,s)} \alpha(r) \alpha(s), \]

where for any region $t$,

\[ \alpha(t) = \frac{1}{\left(1 + \mbox{cov}(t)\right) \mbox{freeVol}^4(t)}, \]

$\mbox{sel}(r,s)$ is the number of times $r$ and $s$ have been part of a lead or selected for exploration, $\mbox{conn}(r,s)$ estimates the progress made by the low-level planner in extending the tree from $r$ to $s$, $\mbox{cov}(t)$ estimates the tree coverage of the region $t$, and $\mbox{freeVol}(t)$ estimates the free volume of $t$. Additional edge cost factors can be added with the addEdgeCostFactor() function, and Syclop's list of edge cost factors can be cleared using clearEdgeCostFactors() .

typedef boost::function< void(int,
int, std::vector< int > &)> 
LeadComputeFn
 Leads should consist of a path of adjacent regions in the decomposition that start with the start region and end at the end region. Default is $A^\ast$ search.

Public Member Functions

 Syclop (const SpaceInformationPtr &si, const DecompositionPtr &d, const std::string &plannerName)
 Constructor. Requires a Decomposition, which Syclop uses to create high-level leads.
ompl::base::Planner Interface
virtual void setup ()
 Perform extra configuration steps, if needed. This call will also issue a call to ompl::base::SpaceInformation::setup() if needed. This must be called before solving.
virtual void clear ()
 Clear all internal datastructures. Planner settings are not affected. Subsequent calls to solve() will ignore all previous work.
virtual base::PlannerStatus solve (const base::PlannerTerminationCondition &ptc)
 Continues solving until a solution is found or a given planner termination condition is met. Returns true if solution was found.
Tunable parameters
void setLeadComputeFn (const LeadComputeFn &compute)
 Allows the user to override the lead computation function.
void addEdgeCostFactor (const EdgeCostFactorFn &factor)
 Adds an edge cost factor to be used for edge weights between adjacent regions.
void clearEdgeCostFactors ()
 Clears all edge cost factors, making all edge weights equivalent to 1.
int getNumFreeVolumeSamples () const
 Get the number of states to sample when estimating free volume in the Decomposition.
void setNumFreeVolumeSamples (int numSamples)
 Set the number of states to sample when estimating free volume in the Decomposition.
double getProbShortestPathLead () const
 Get the probability [0,1] that a lead will be computed as a shortest-path instead of a random-DFS.
void setProbShortestPathLead (double probability)
 Set the probability [0,1] that a lead will be computed as a shortest-path instead of a random-DFS.
double getProbAddingToAvailableRegions () const
 Get the probability [0,1] that the set of available regions will be augmented.
void setProbAddingToAvailableRegions (double probability)
 Set the probability [0,1] that the set of available regions will be augmented.
int getNumRegionExpansions () const
 Get the number of times a new region will be chosen and promoted for expansion from a given lead.
void setNumRegionExpansions (int regionExpansions)
 Set the number of times a new region will be chosen and promoted for expansion from a given lead.
int getNumTreeExpansions () const
 Get the number of calls to selectAndExtend() in the low-level tree planner for a given lead and region.
void setNumTreeExpansions (int treeExpansions)
 Set the number of calls to selectAndExtend() in the low-level tree planner for a given lead and region.
double getProbAbandonLeadEarly () const
 Get the probability [0,1] that a lead will be abandoned early, before a new region is chosen for expansion.
void setProbAbandonLeadEarly (double probability)
 The probability that a lead will be abandoned early, before a new region is chosen for expansion.

Protected Member Functions

virtual MotionaddRoot (const base::State *s)=0
 Add State s as a new root in the low-level tree, and return the Motion corresponding to s.
virtual void selectAndExtend (Region &region, std::vector< Motion * > &newMotions)=0
 Select a Motion from the given Region, and extend the tree from the Motion. Add any new motions created to newMotions.
const RegiongetRegionFromIndex (const int rid) const
 Returns a reference to the Region object with the given index. Assumes the index is valid.

Protected Attributes

int numFreeVolSamples_
 The number of states to sample to estimate free volume in the Decomposition.
double probShortestPath_
 The probability that a lead will be computed as a shortest-path instead of a random-DFS.
double probKeepAddingToAvail_
 The probability that the set of available regions will be augmented.
int numRegionExpansions_
 The number of times a new region will be chosen and promoted for expansion from a given lead.
int numTreeSelections_
 The number of calls to selectAndExtend() in the low-level tree planner for a given lead and region.
double probAbandonLeadEarly_
 The probability that a lead will be abandoned early, before a new region is chosen for expansion.
const SpaceInformationsiC_
 Handle to the control::SpaceInformation object.
DecompositionPtr decomp_
 The high level decomposition used to focus tree expansion.
RNG rng_
 Random number generator.

Detailed Description

Synergistic Combination of Layers of Planning.

Short description
Syclop is a multi-layered planner that guides a low-level sampling-based tree planner through a sequence of sequence of discrete workspace regions from start to goal. Syclop is defined as an abstract base class whose pure virtual methods are defined by the chosen low-level sampling-based tree planner.
External documentation
E. Plaku, L.E. Kavraki, and M.Y. Vardi, Motion Planning with Dynamics by a Synergistic Combination of Layers of Planning, in IEEE Transactions on Robotics, 2010. DOI: [10.1109/TRO.2010.2047820](http://dx.doi.org/10.1109/TRO.2010.2047820)

Definition at line 70 of file Syclop.h.


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