search

uzh-rpg / rpg_quadrotor_control

Quadrotor control framework developed by the Robotics and Perception Group
Other
585 stars 185 forks source link

SVO 2.0 as odom feedback #65

Open mrsp opened 6 years ago

mrsp commented 6 years ago

Hi

Digging up a bit in your code, your feedback is based on an instance of "state_predictor", have you tried using your SVO 2.0 as your odometry feedback?

Thx in advance

mfaessle commented 6 years ago

The state_predictor is only responsible to compensate for latencies of the state estimate and the resulting control command. When the state estimate arrives at the controller, it already has a certain age from which it is predicted forward until the current time. Similarly, we can consider delays of the resulting control command, e.g. when using a wireless link to send commands, by integrating the state estimate forward based on previously applied control commands.

The origin of the state estimate does not matter as long as it is passed to the controller as a nav_msgs/Odometry message. And of course we have used SVO 2.0 together with the Multi Sensor Fusion framework (https://github.com/ethz-asl/ethzasl_msf) to get a state estimate. This is how all our basic demos work, e.g. https://youtu.be/7-kPiWaFYAc