rlsutton1My agenda this week, we'll see how much I get done!
| Issue | Intended Solution |
|---|---|
| Poor adherence to planned path. | Calculate the curvature of the bezier slightly in front of the robots current position. |
| Poor adherence to planned path. | Convert route gradient to a fixed path. |
| Poor detection of objects via depth camera. | Come up with or change to a new algorithm for locating edges in the point cloud. |
| Erratic turning. | Stabilize the turn radius via making adjustments to the radius over half a second period at 100ms intervals, |
| Slow reaction time to new obstacles. | Run the route planner on demand when there is a change detected in the environment rather than on a set frequency. |
| Inconsistent perception of position and size of objects detected via the point cloud. This is in part because of the broken Intel Realsense integration to OpenNI2 and my inability to compensate for it using the available information provided by the OpenNI2 API. | Rather than track via an occupancy grid, move to tracking objects and noting their size and location. Subsequent observation would update the size and location for an object with a suitably near match of location and size, thus over time improving the accuracy of the perceived size and location of the object. |
| Update the gallery picture to a more current build of the robot | Self explanatory really. |
Any tips, suggestions or questions gratefully received.
Discussions
Become a Hackaday.io Member
Create an account to leave a comment. Already have an account? Log In.