Differences

This shows you the differences between two versions of the page.

--- software:cram:tutorials:pick-and-place-primer [2013/03/26 12:33] – [Executing Robot Plans] winkler
+++ software:cram:tutorials:pick-and-place-primer [2016/05/19 09:19] (current) – external edit 127.0.0.1
@@ Line 11: / Line 11: @@
 ==== Prerequisites ====
-It is assumed that you successfully completed the [[software:cram:setup|installation of the CRAM system]] before starting this lesson.
+It is assumed that you successfully completed the [[software:cram:setup:auto-install|installation of the CRAM system]] before starting this lesson.
 ==== Getting the Code ====
 To download the lesson, in your ROS workspace, do the following:
-<code>
+<code bash>
 $ rosws set lesson01 --git https://github.com/ai-seminar/cram-tutorials.git
 $ rosws update lesson01
+$ source ~/.bashrc # To reread the available ROS packages in the package path
 </code>
 If you're not using rosws, although it is strongly recommended, you can get the repository [[https://github.com/ai-seminar/cram-tutorials|here]].
@@ Line 23: / Line 24: @@
 ==== Preparation ====
 To make sure that all software modules are built correctly and all dependencies are met, do
-<code>
+<code bash>
 $ rosmake lesson01
 </code>
@@ Line 30: / Line 31: @@
 ==== Getting started ====
 First, start the Slime REPL. This is the command line from which you will execute Lisp code and control the CRAM system. If you get lost when using Emacs, [[http://refcards.com/docs/wingb/xemacs/xemacs-refcard-a4.pdf|this cheat sheet]] might help you.
-<code>
+<code bash>
 $ rosrun roslisp_repl repl
 </code>
-Now, load the lesson code. In order to do so, press ',' when the `CL-USER>` prompt shows up. After that, enter:
+Now, load the lesson code. In order to do so, press '','' when the ''CL-USER>'' prompt shows up. After that, enter:
 <code>
 Command: ros-load-system
@@ Line 46: / Line 47: @@
 ROS package `lesson01', System (default `lesson01'):
 </code>
-by just pressing Return again. This will load the ASDF system `lesson01` from within the ROS package `lesson01`. The REPL guesses the correct default system here because there is only one system in the .asdf file. When loading the system for the first time, a lot of Lisp system dependencies are compiled. This can take quite some time.
+by just pressing Return again. This will load the ASDF system ''lesson01'' from within the ROS package ''lesson01''. The REPL guesses the correct default system here because there is only one system in the .asdf file. When loading the system for the first time, a lot of Lisp system dependencies are compiled. This can take quite some time.
 If everything went well, you should see something similar to the following:
 {{ :software:cram:tutorials:repl_after_load.png?800 |}}
-To get into the correct lesson package, press ',', enter '!p', press Return and enter:
+To get into the correct lesson package, press '','', enter ''!p'', press Return and enter:
 <code>
 Package: LESSON01
@@ Line 57: / Line 58: @@
 From now on, you will have a prompt like
-<code>
+<code lisp>
 LESSON01>
 </code>
@@ Line 64: / Line 65: @@
 === Starting the simulation environment ===
 Now, back in the terminal, start
-<code>
+<code bash>
 $ roslaunch lesson01 lesson01.launch
-</code>(with-designators
+</code>
-                  ((mug (object `((desig-props:name "mug_1")))))
-                (cram-plan-library:perceive-object 'cram-plan-library:the mug)
+This will start Gazebo, spawn a PR2 robot and a table with two mugs on it.
+{{ :software:cram:tutorials:gazebo_lesson01_spawn.png?800 |}}
+The PR2 and the mugs are dynamic objects while the table is fixed to the ground, i.e. you can't move it. (If you encounter problems with parts of the simulation not showing up correctly, see [[pick-and-place-primer# Pitfalls and known issues |here]].)
-This will start Gazebo, spawn a PR2 robot and a table with two mugs on it.{{ :software:cram:tutorials:gazebo_lesson01_spawn.png?800 |}}
-The PR2 and the mugs are dynamic objects while the table is fixed to the ground, i.e. you can't move it.
 In yet another terminal, open rviz:
-<code>
+<code bash>
 $ rosrun rviz rviz
 </code>
-This will show you the PR2 model and the bottom map [[software:cram:tutorials:general:rviz-config|if properly configured]]:Node name is
+This will show you the PR2 model and the bottom map [[software:cram:tutorials:general:rviz-config|if properly configured]]:
 {{ :software:cram:tutorials:rviz_lesson01_spawn.png?800 |}}
-Together with Gazebo, an instance of the [[http://ros.org/wiki/rosmaster|ROS master]] (`roscore`) was started when executing the roslaunch command. All components using ROS connect to this master and exchange data through it's messaging system. In order to connect our REPL to it, enter the following command in the Slime REPL:
+Together with Gazebo, an instance of the [[http://ros.org/wiki/rosmaster|ROS master]] (''roscore'') was started when executing the roslaunch command. All components using ROS connect to this master and exchange data through it's messaging system. In order to connect our REPL to it, enter the following command in the Slime REPL:
-<code>
+<code lisp>
 LESSON01> (start-lesson)
 </code>
 You will see output similar to this:
-<code>
+<code lisp>
 [(LESSON01) INFO] 1364288497.123: Starting lesson 01.
 [(ROSLISP TOP) INFO] 1364288497.331: Node name is /cram_hl_316821_1364288497
@@ Line 103: / Line 104: @@
 [(ROSNODE) INFO] 445.747: ROS init #<FUNCTION PR2-MANIPULATION-PROCESS-MODULE::INIT-PR2-MANIPULATION-PROCESS-MODULE>.
 </code>
-The second line, starting with "Node name is" includes the ROS node name of your current CRAM node instance. You will need this later.
+The second line, starting with ''Node name is'' includes the ROS node name of your current CRAM node instance. You will need this later.
 To try out whether the components are connected properly, you can use the Gazebo/CRAM interface to get the pose of one of the mugs:
-<code>
+<code lisp>
 LESSON01> (cram-gazebo-utilities:get-model-pose "mug_1")
 </code>
 which should now yield the pose for the mug in question:
-<code>
+<code lisp>
 #<POSE-STAMPED
    FRAME-ID: "map", STAMP: 0.0
@@ Line 119: / Line 120: @@
 ==== The Semantic Environment ====
-The given lesson environment includes a simplistic semantic map for the scene. It includes the table which is placed in front of the PR2 robot. In terms of the semantic map, this table is called "//popcorn_table//" and can be referred to from within CRAM. When rviz is running, add a new display entity of type "//Marker//".
+The given lesson environment includes a simplistic semantic map for the scene. It includes the table which is placed in front of the PR2 robot. In terms of the semantic map, this table is called ''popcorn_table'' and can be referred to from within CRAM. When rviz is running, add a new display entity of type ''Marker''.
 {{ :software:cram:tutorials:select_marker.png?200|}}
-This display type can visualize ROS messages of type "//visualization_msgs::Marker//", which are boxes, spheres, meshes, and other 3D object types in different colors. As "Marker Topic", you choose the node name of the current CRAM instance you initiated from within the Slime REPL, as mentioned earlier., and select the topic "semantic_map_markers" under it.
+This display type can visualize ROS messages of type ''visualization_msgs::Marker'', which are boxes, spheres, meshes, and other 3D object types in different colors. As ''Marker Topic'', you choose the node name of the current CRAM instance you initiated from within the Slime REPL, as mentioned earlier, and select the topic ''semantic_map_markers'' under it.
 {{ :software:cram:tutorials:choose_cram_node_rviz.png?200|}}
 You can now visualize the semantic map using the following code from the REPL:
-<code>
+<code lisp>
 LESSON01> (sem-map-coll-env:publish-semantic-map-markers)
 </code>
@@ Line 132: / Line 133: @@
 {{ :software:cram:tutorials:sem_map_published.png?400 |}}
-Adding another display entity of type "//CollisionMap//" and choosing the topic "/kipla/location_costmap" as the "Topic" parameter enables rviz to visualize the output of location costmap related designator resolutions:
+Adding another display entity of type ''CollisionMap'' and choosing the topic ''/kipla/location_costmap'' as the ''Topic'' parameter enables rviz to visualize the output of location costmap related designator resolutions:
-<code>
+<code lisp>
 LESSON01> (top-level
             (with-designators ((loc (location `((desig-props:on Cupboard)
@@ Line 142: / Line 143: @@
 {{ :software:cram:tutorials:location_costmap_displayed.png?400 |}}
 and a pose on the table which was drawn from the shown samples, i.e.
-<code>
+<code lisp>
 #<POSE-STAMPED
    FRAME-ID: "map", STAMP: 0.0
@@ Line 155: / Line 156: @@
   * object-in-hand
-The function //perceive-object// transforms the vague description of the object to pick up into a 3D pose information. Using the //gazebo_perception_process_module//, the effective model state of the simulated object is extracted and transformed into an object pose. All of this happens internally and can be triggered using:
+The function ''perceive-object'' transforms the vague description of the object to pick up into a 3D pose information. Using the ''gazebo_perception_process_module'', the effective model state of the simulated object is extracted and transformed into an object pose. All of this happens internally and can be triggered using:
-<code>
+<code lisp>
 LESSON01> (top-level
+            (move-away-arms)
             (with-process-modules
               (with-designators
@@ Line 164: / Line 166: @@
 </code>
 This results in an object designator description defining the object in question:
-<code>
+<code lisp>
 #<OBJECT-DESIGNATOR ((AT
                       #<LOCATION-DESIGNATOR
@@ Line 176: / Line 178: @@
 </code>
-The //object-in-hand// function is actually a goal the robot tries to achieve. It is called using
+The ''object-in-hand'' function is actually a goal the robot tries to achieve. It is called using
-<code>
+<code lisp>
 LESSON01> (top-level
+            (move-away-arms)
             (with-process-modules
               (with-designators
@@ Line 189: / Line 192: @@
 {{ :software:cram:tutorials:location_costmap_drive.png?400 |}}
-On the lesson code (specifically in `lesson01.lisp`), there is a function for executing both parts in order:
+In the lesson code (specifically in ''lesson01.lisp''), there is a function for executing both parts in order:
-<code>
+<code lisp>
 (defun get-mug-1 ()
   ...)
@@ Line 196: / Line 199: @@
 You can call this function by simply starting it from the REPL:
-<code>
+<code lisp>
 LESSON01> (get-mug-1)
 </code>
+It will also take care of functions like moving the robot's spine up all the way and getting the arms out of the way before doing navigation/manipulation tasks.
 The PR2 will drive around and will try to get into a suitable position for grasping. It then will try to grasp the object.
+Once the object is being grasped, the robot will execute a motion to achieve the following grasp and carrying pose:
+{{ :software:cram:tutorials:pr2_mug_carry.png?370| }}
+{{ :software:cram:tutorials:pr2_grasp_mug.png?300 |}}
+=== Handle-based grasping ===
+The fact that the PR2 grasps the mug on it's handle results from two things:
+  * An extensive reasoning machinery that checks for the existance of handles on an object to grasp, plus an algorithm that calculates the respective grasp poses
+  * The definition of the object handle in the source code
+The system relies on information given to it about handles. This information if thought to be stored in an external knowledge source such as [[software:knowrob|KnowRob]] but can be given by hand here.
+The important part in the source code for handle definition is found here:
+<code lisp>
+(with-designators
+    ((handle-loc (location
+                  `((desig-props:pose
+                     ,(tf:make-pose
+                       (tf:make-3d-vector
+                        -0.12 0.0 0.07)
+                       (tf:euler->quaternion
+                        :ax (/ pi 2)))))))
+     (mug-handle (object
+                  `((desig-props:type desig-props:handle)
+                    (desig-props:at ,handle-loc))))
+     (mug (object `((desig-props:name "mug_1")
+                    (desig-props:handle ,mug-handle)))))
+</code>
+The variable ''handle-loc'' defines the relative pose of the handle with respect to the bottom-most center point of the mug, i.e. the pose that is returned by the ''gazebo_perception_process_module''.
+''mug-handle'' then defined a handle-object that is then placed as a handle into the actual mug-object, ''mug''.
+==== Completing the Pick and Place Task ====
+To make the robot execute the task at hand, the function ''displace-mug-1'' has been prepared. You can find it in ''lesson01.lisp''. It is started by:
+<code lisp>
+LESSON01> (displace-mug-1)
+</code>
+During the runtime of this function, the robot will try to get hold of the object (i.e., pick it up) and then put it down on a different spot of the table.
+Since finding a good spot to put the object down on the table isn't that easy, it is possible that the PR2 will drive around a bit and try different spots before it finds a suitable location.
+For putting down the object, the same techniques with respect to location costmaps are used as in the pick-up part of this tutorial.
+==== Pitfalls and known issues ====
+In case the table doesn't show up completely when starting the lesson launchfile (and, Gazebo in particular), you have to restart the Gazebo client GUI. The problem will show you a scene similar to this one:
+{{ :software:cram:tutorials:issue_pr2_table_nospawn.png?300 |}}
+Just close the Gazebo window and restart the client, using:
+<code bash>
+$ `rospack find gazebo`/gazebo/bin/gzclient -g `rospack find gazebo`/lib/libgazebo_ros_paths_plugin.so
+</code>
+The (simulated) table is physically there, but the visual is missing. The simulation could go on without it, but for the scenery's sake, you can get it back into the scene with this fix.
+ --- //[[|Winkler, Jan]] 2013/03/28 11:10//