Documentation Status

motion_planners: chomp_motion_planner | interpolated_ik_motion_planner | ompl | ompl_ros_interface | sbpl

Package Summary

Documented

A library of search-based planning algorithms

Package Summary

Released No API documentation

Search-based Planning Library (SBPL)

  • Maintainer: E. Gil Jones <gjones AT willowgarage DOT com>
  • Author: Max Likhachev <maxim AT cs.cmu DOT edu>
  • License: BSD

Package Summary

Released No API documentation

Search-based planning library (SBPL).

  • Maintainer: Ioan Sucan <isucan AT willowgarage DOT com>
  • Author: Search-Based Planning Lab
  • License: BSD
  • External website: http://http://sbpl.net

Package Summary

Released No API documentation

Search-based planning library (SBPL).

  • Maintainer status: maintained
  • Maintainer: Michael Ferguson <mike AT vanadiumlabs DOT com>
  • Author: Search-Based Planning Lab
  • License: BSD
  • External website: http://sbpl.net

Package Summary

Released No API documentation

Search-based planning library (SBPL).

  • Maintainer status: maintained
  • Maintainer: Michael Ferguson <mike AT vanadiumlabs DOT com>
  • Author: Search-Based Planning Lab
  • License: BSD
  • External website: http://sbpl.net
Cannot load information on name: sbpl, distro: kinetic, which means that it is not yet in our index. Please see this page for information on how to submit your repository to our index.
Cannot load information on name: sbpl, distro: lunar, which means that it is not yet in our index. Please see this page for information on how to submit your repository to our index.

Please note that the source link above is wrong. Please download the SBPL here:

This package implements a generic set of motion planners using search based planning. It was developed by Maxim Likhachev at the University of Pennsylvania in collaboration with Willow Garage.

Video

Documentation

The API documentation for this package can be found here.

A good resource for technical information on the planners in the SBPL package is Maxim Likhachev's presentation (pdf,movie) at the ROS Cotesys School at TUM, Germany in November 2010.

Motion Primitives

Motion primitives are short, kinematically feasible motions which form the basis of movements that can be performed by the robot platform. Search-based planners (like the ones in this library) can generate paths from start to goal configurations by combining a series of these motion primitives. The result is a smooth kinematically feasible path for the robot to follow.

Several planners make use of SBPL and these motion primitives to produce smooth paths:

  • sbpl_arm_planner - 3DoF and 6DoF arm planning for the PR2

  • sbpl_lattice_planner - (x,y,yaw) planning for robot navigation (handles non-circular footprints and nonholonomic constraints)

Pre-made Motions (x,y,yaw navigation)

This package has two pre-made motion primitive files for the PR2 (though these motions are a reasonable choice for many other robots).

  • matlab/mprim/pr2.mprim - Motion primitives that only allow forward, backward, turn-in-place, and forward-and-turn-arc motions
  • matlab/mprim/pr2sides.mprim - Same as pr2.mprim (previous) but includes strafing left and right as allowable motions

Generating your own Motions (x,y,yaw navigation)

Sometimes the pre-made will not be good enough either because you are using a map of a different resolution (a motion primitive file only works on a map with its corresponding resolution) or because your robot's motions don't match with any of the pre-made motions.

In the matlab/mprim folder there are several matlab scripts which are easy to modify and generate new motion primitives. They are named according to the types of motions they generate.

  • genmprim_unicycle.m - forward, backward, turn-in-place, forward-turn-arcs
  • genmprim_unicycleplussideways.m - like previous but with strafing left/right
  • genmprim_unicycleplussidewaysplusbackturn.m - like previous but with backward-turn-arcs
  • genmprim_unicycleplussidewaysplusbackturnplusdiag.m - like previous but with diagonal moves

Parameters

  • Resolution - Near the top of each of these files there is a "resolution" variable which you can set you your map's resolution.
  • Number of angles - The variable "numberofangles" represents the yaw discretization. The more you have the better your paths will look but the planning time will also increase because there are more states to search. Be careful changing this as you will need to add/remove motion vectors in the rest of the script.
  • Number of motions - The variable "numberofprimsperangle" represents the number of motions
  • Multipliers - There is a multiplier for each type of motion (forwardcostmult, backwardcostmult, etc) which is multiplied by the cost of using the motion. You can use these multipliers to discourage certain motions (making the multiplier larger) so that the robot only uses them if it has to.
  • Motions - There will be several vectors (each for a different direction) of motions. Each element of a vector is a 4-tuple representing (dx, dy, dyaw, multiplier). You can change what multiplier a motion uses and how it moves by changing these. Remember that the number of 4-tuples in each vector must equal the "numberofprimsperangle" variable.

ROS API

  • The ROS API is UNREVIEWED and UNSTABLE
  • The C++ API is UNREVIEWED and UNSTABLE

Wiki: sbpl (last edited 2013-05-31 20:13:37 by BenCohen)