ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

robot_localization: adapt internal motion model.

asked 2016-03-16 03:47:51 -0600

jbler gravatar image

updated 2016-03-17 09:07:34 -0600

Hi folks,

  1. I want to use the robot_localization package with my parrot AR.Drone 2. I previously did some testing within the matlab enviroment and created a simplified quadcopter model (in matlab I get reasonable results). Therefore my thoughts was to adapt the internal motion model of the predict code to my quadcopter model. Is the code designed to perform changes in the internal motion model?

  2. My aim of using the robot_localization package is to predict long time ahead (around 120ms) to compensate communication delays. Therefore I will also need to add control inputs to achieve a accurate prediction of the state of my drone. AFAIK there is no possibility to use control inputs, does anyone know a good workaround?

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted

answered 2016-05-08 20:19:55 -0600

Tom Moore gravatar image

Wow, apologies for the delay in responding to this.

I've used r_l with the AR.Drone 2.0 before, and it worked fine. In fact, one of the test bag files that is in the r_l test set in the repo is from that drone.

Having said that, if you find the model isn't accurate enough, I want to move to a plugin-based architecture so that we can support other kinematic models.

Also, check out what's in the repo on the indigo-devel branch now. I've added a control term. I've also just merged a great PR that lets the filter handle lagged data.

edit flag offensive delete link more

Question Tools

1 follower


Asked: 2016-03-16 03:47:51 -0600

Seen: 391 times

Last updated: May 08 '16