SSPP

From CAS-Robotics

Contents 1 SSPP (Search Space Path Planner) driver for player/stage 1.1 Download 1.2 Compilation and Usage 1.3 Specifications / Operation 1.4 Configuration Parameters 1.5 Examples 1.6 Stage Test

SSPP (Search Space Path Planner) driver for player/stage

This is a Path Planning driver for online path planning using an offline generated Search Space. It's a very efficient method for path planning in a mapped environment with tight spaces. The Path Planning takes into consideration the size of the robot and generates a search space of the environment that can be used for online Path Planning.It's based on an efficient sampling method and collision detection that makes path planning for tight and narrow environments possible and fast. I would appreciate any feedback, recommendations and bug reporting that can lead to improving the performance of this driver.

Download

• Download

Compilation and Usage

To compile the driver successfully, you should:

• Have pkg-config pointing correctly to the player/stage.
• Use the make file script to compile the driver and generate the shared object "SSPP.so"

make

To install:

• Copy SSPP.so to /usr/local/lib/ or
• Create a symbolic link where you're running your configuration file

ln -s SSPP_PATH/SSPP.so

where SSPP_PATH is the path where the SSPP driver is extracted and compiled

Specifications / Operation

The driver requires:

• A position device "position2d" that can provide location estimation (AMCL usually).
• A position device "position2d" that the driver can control (usually VFH or nd).
• A map device "map" that will provide an occupancy grid of the environment.

The driver provides:

• A Planner interface for accepting target goals and sending WayPoints.

Handled requests:

• PLAYER_PLANNER_REQ_GET_WAYPOINTS
• PLAYER_PLANNER_REQ_ENABLE
• PLAYER_PLANNER_CMD_GOAL
• PLAYER_PLANNER_DATA_STATE

On start, the driver will read the map from the map interface and generate the search space using the parameters specified in the configuration file (occupied and unknown cells are considered occupied). This step takes some time and the generated search space will be saved in a file so that the next time the driver is restarted it doesn't have to regenerate the space but load it from the file.

When the driver receives a new goal, it finds the path from the current robot's location to that goal and save the path's intermediate steps as waypoints and send them to the client once requested. If the driving is enabled (from the client side) then the driver finds the waypoint on the generated path that is "traversableDist" away from the current location and sends this waypoint as a goal to the underlying position device (nd driver is prefered but it can also be VFH+). The driver will keep on checking how far is the robot from the goal and stops it when it reaches the destination or when the driving is disabled from the client side.

Configuration Parameters

- traversableDist (float) 
	- Default: 1.0
	- Specify how far you want your waypoint be from your current location. This helps in 
	  reducing the local minima case.
- narrowestPassage (float)
	- Default: 0.987
	- Based on the map you specified, what is the narrowest passage that you want the robot to
	  be able to pass through. This is used in determining the expansion radius used for the 
	  collision detection.
- distanceToGoal (float) 
	- Default: 2
	- Define how close you want to be to the goal destination.
- bridgeLen (float)
	- Default: 2
	- Specifies the length of the segments used for the bridge test.
- bridgeRes (float) 
	-Default: 0.1
	- Defines how dense should the sampling resolution be while using the bridge test.
- regGridLen (float) 
	-Default: 0.2
	- Defines how dense should be the regular sampling resolution be.
- regGridConRad (float)
	- Default: 0.4
	- Define what radius should be used for connecting regular grid samples.
- obstPenalty (float)
	- Default: 3.0
	- Defines the how close to an obstacle should we approach to start penalizing.
- bridgeConRad (float) 
	- Default: 0.5
	- Specifies the radius for connecting samples generated using the bridge test.
- debug (bool)
	- Default: false
	- Specify if you want to debug the driver or not. This will save the generated space
	  and the path into an image file in the local folder.

Examples

driver
(
  name "SSPP"
  plugin "SSPP.so"
  provides ["planner:0"]
  requires ["output:::position2d:1" "input:::position2d:2" "map:0"] 
  traversableDist 1
  narrowestPassage 0.9
  distanceToGoal 0.2
  bridgeLen 3.0
  bridgeRes 0.2
  regGridLen 0.4
  regGridConRad 0.4
  bridgeConRad 0.5
  obstPenalty 2
)

Stage Test

• The directory stage_test contains a stage configuration for testing SSPP. Simply run:

player sspp.cfg

• You can also use playernav to debug it and send goal commands.

driver
(
  name "stage"
  plugin "libstageplugin"
  provides ["simulation:0"]
  worldfile "cas.world"
)
driver
(
  name "mapfile"
  provides ["map:0"]
  filename "casarea_sub.jpeg"
  resolution 0.05  
)
driver
(
  name "stage"
  model "robot"
  provides ["position2d:0" "laser:0"]
)
driver
(
  name "nd"
  provides ["position2d:1"]
  requires ["output:::position2d:0" "input:::position2d:0" "laser:0"]
  max_speed [0.3 30.0]
  min_speed [0.1 10.0]
  goal_tol [0.3 15.0]
  wait_on_stall 1

  rotate_stuck_time 5.0
  translate_stuck_time 5.0
  translate_stuck_dist 0.15
  translate_stuck_angle 10.0

  avoid_dist 0.4
  safety_dist 0.0
  laser_buffer 1
  alwayson 1
)
driver
(
  name "amcl"
  update_thresh [0.05 0.175] # 10 degrees angle and 5 cm
  provides ["localize:0" "position2d:2"]
  requires ["odometry:::position2d:0" "laser:0" "laser:::map:0"]
)
driver
(
  name "SSPP"
  plugin "SSPP.so"
  provides ["planner:0"]
  requires ["output:::position2d:1" "input:::position2d:2" "map:0"]
)