Skip to content

Latest commit

 

History

History
71 lines (32 loc) · 2.02 KB

DOCUMENTATION.md

File metadata and controls

71 lines (32 loc) · 2.02 KB
Raw

VL53L1X Scanner Code Documentation

UNDER CONSTRUCTION!

1. Adaptive Scan Resolution

Optimizes angle between scans based on the previous single measurement.

Pros:

  • Is good for real-time operations (i.e. localization)
  • Gives results faster
  • Gives less data but with similar information

Cons:

  • Outputs PointCloud2. Needs resampling for LaserScan message.
  • May miss some measurements on edges of an object.

Math:

(1) Finding theta. Assuming that two scans will have same measurements. Applying cosines theorem.

adaptive_res_1

(2) In boundaries of the hardwares, theta = f(dist) has reciprocal distribution features. After removing the effect of this property, applying line fitting to find coefficients. See /scripts/adaptive_scan_resolution_find_coefficients.py

adaptive_res_2

(3) Line fitting result of theta = p(1/dist)

adaptive_res_3

Note:

Actually, I have tried predicting orientation of the wall to guess the direction of next measurement, but it failed. To explain it, firstly, rapid changes in the measurements caused fluctuations in guessed orientation of walls. Secondly, I have solved this issue with applying exponentially weighted averages on to the guess of wall orientation, but this caused overshoot at the end of walls. As a result, I have quit working with orientation predictions and came with this idea which is simpler and faster.

2. PointCloud Calculation

  • Math calculation for measurement to pointcloud conversion:

pointcloud_calculation

Notes

  • Stepper motor delay set as 2.25ms instead of 2ms, because it was missing steps.

  • (TODO) 28BYJ-48 steps per revolution is 2048 instead of 2038.

  • Intensity value is equal to signal_rate (This may change in the future)

TODO:

  • 3d best quality/speed parameters
  • 2d best quality/speed parameters
  • 2d select scan ring configuration