In autonomous exploration considering the pose uncertainty, is there any way to predict before hand the pose uncertainty the robot will have if it takes a particular action. This is to plan motion to increase the localizability while making exploration decision.
Hi Hari
You will find the answer to your question (at least one answer) in this paper:
Uncertainty-aware Receding Horizon Exploration and Mapping Using Aerial Robots
Video describing the solution:
https://www.autonomousrobotslab.com/autonomous-navigation-and-exploration.html
The software is open source (rely on ROS):
https://github.com/unr-arl/rhem_planner
The software has been developed for a quadrotors (6 degrees of freedom) but I guess it can be easily adapted to 3 DOF in case your robot is a wheeled robot.
Fabrice
When working with navigation, there a lot of ways to study uncertainty.
If you see your localization problem is a Gaussian distribution, the Kalman filter is a good choice. You can also try Particle filter.
Hi all,
Thank a lot for the replies. I tried a method also taking some of your suggestions. I did ray tracing to obtain simulated laser scan over the existing grid map and then used a laser beam model to find out p(z|x). However I'm bit confused about the final step of estimating the pose with covariance since I did not use a particle filter.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence