ras_cal's profile - activity

Our indoor wheeled robot with wheel encoders has an IMU without magnetometers. At startup in a random orientation the yaw output is 0. We are able to follow a wall and cross into another room where an internal GPS is set up to determine X, Y and yaw with respect to global map frame of the room. We are using robot_localization for fusion, but are unsure how to switch to the room map without causing a discontinuity in the localization and therefore in the motion. Is it possible to go from map1 --> odom to map2 --> odom ?

Indoor Robot with No absolute starting yaw Our indoor wheeled robot with wheel encoders has an IMU without magnetometers

we used a path tracking algorithm. In our case Pure Pursuit worked, although we had to develop from scratch.

We couldn't find anything that could be used with only minor modifications so we developed a pure pursuit type of naviga

I've had this issue. Seems to be related to graphics card. Take a look at http://answers.ros.org/question/44818... and http://answers.ros.org/question/22382...

mathworks link

Thanks. How would the local_planner know to not stop when it reaches each intermediate goal?

This question was asked by someone else in 2013, but never answered. So giving it another shot as I'm sure there are other people besides me with similar question. Using ROS Indigo, we have an indoor robot that is to follow a pre-planned trajectory at a constant speed. There are no obstacles. Can move_base, base_local_planner be used to generate the required velocity commands such that the robot passes through the goal (waypoint) at a predetermined pose and speed and continues towards the next goal at the same linear speed and without stopping until it reaches the last (end) goal? If move_base won't work is there a package that could?

The yocs_waypoints_navi package seems to use move_base and internally creates a new goal before the old one is reached for "waypoint navigation", but there is very little documentation on how to use.

I think the global planner goal format is a geometry_msgs/PoseStamped message which is just a point and pose. I would need a goal with a format that includes a point (x,y) as well as a linear speed vector (vx, vy) defined for each waypoint.

by changing the base_link_frame parameter value from base_link to base_footprint I was able to connect the odom frame to base_footprint. I can now run the robot and visualize the ekf_localization output in RViz. Interestingly, the base_footprint to base_link transform is now broadcast by /robot_state_publisher

Thank you. I've posted image.

I have left the robot_localization odom_frame parameter with value "odom" and base_link_frame to "base_link" defaults. The tf tree, while running robot_localization and the robot driver ca_driver with publish_tf set to "false". Looks like this (apparently I can't upload a file with less than 5 points): odom floating. base_footprint ---> base_link (broadcaster /ekf_localization) base_link --> right_wheel link (broadcaster /robot_state_publisher) base_link --> left_wheel link (broadcaster /robot_state_publisher) ....

why is ekf_localization broadcasting base_footprint --> base_link ? and leaving odom floating? (not connected to anything) I can't even find base_footprint anywhere in the robot_localization parameter list. It is part of the robot description. Any help clarifying this would be appreciated.

Edit

Adding image of original tf tree

I am using ROS Indigo on a differential drive robot for indoor use (no GPS). The path is pre-determined and the velocity along the path is to be controlled. I believe that move_base node would be overkill and probably would not work in this case. Speed along path needs to be at a certain rate and if an obstacle is detected the unit will stop and send an alarm. I think a 2-D PID speed controller (linear and angular vel.) and a path planner that reads from the pre-planned path and provides the commanded 2-D speed is all that is needed. Is there an existing ROS node I could use or modify?

Using ROS Indigo and the create_autonomy package I'm able to control an iRobot Create 2. I would like to calibrate the odometry using a method similar to that used for the Turtlebot - or simpler. Just by moving manually a certain distance and measuring with a measuring tape the robot distance to obtain a correction factor that can then be incorporated into the driver to improve accuracy. Same for angular correction. I'm aware of http://answers.ros.org/question/23252... but would rather measure and enter a linear and angular calibration correction factor for a given floor type that takes care of any inconsistencies. Any help or insight would be appreciated.

Was able to solve by following the steps here http://answers.ros.org/question/20320... The key was deleting the ws_A build and devel folders. also insured that roscore wasn't running - not sure if that mattered or not.

I have two workspaces ws_A and ws_B with same packages. ws_A is the original and ws_B is a copy. I have gone through the catkin and overlaying tutorials with no avail. I have modified ROS_PACKAGE_PATH, done $ workspace_init, $ catkin_make on ws_B successfully, opened new terminals and done $ source ~/ws_B/devel/setup.bash. Regardless, a roscd package always sends me to the original ws_A folder. I've looked at other accepted answers and tried source /opt/ros/indigo/setup.bash first with no luck. It simply doesn't see ws_B. Using ROS Indigo and Ubuntu 14.04 Any help would be appreciated, in particular the exact steps required to see ws_B.