Skip to content

Latest commit

 

History

History
99 lines (82 loc) · 3.47 KB

TESTCASES.rst

File metadata and controls

99 lines (82 loc) · 3.47 KB
Raw

TESTCASES

(2015-06-10) - transition on curved stairs

Test Criteria:
  • adjustment of beta needs to be smooth
Results:
  • offset on curved stairs in nonlegligable
  • transition from straight to curved stairs leads to matching errors -> play with FoV
  • beta threshold on straight stairs should be around 10°
  • turning point on curved stairs is currently to far behind the wheelchair

(2015-06-09) - test manual climb on straight stairs

Test Criteria:
  • Measure 10 times going up
    • Maximal Distance
    • Stop angle from plot
    • Deviation from straight line
Procedure:
  • Align Wheelchair straight on stairs
  • start service -> kp = 0
  • start bagfile
  • drive backwards
  • measure data
  • drive up and down 10 times
Results:
  • little to no deviation while climbing straight stairs

(2015-06-08) - test automatic climb on curved stairs

Test Criteria:
  • Climb stairs without external alignment help
  • Recognize stairparameters and position on stairs
  • tested on CLA curved stairway
Results:
  • Even tough no extrapolation of the angle was implemented, going up the stairs worked fine.
  • Often a single, scattered measurement point on scan_1 messed up the matching process. -> filter which delets measurements in close proximity of the sensor was implemented
  • Single_Edge_Beta and Double_Edge_Beta performed simalar. Both worked about equally good.
  • Bug of changing beta angle according to direction of travel
    • while going up (dx > 0) kp needed to be positiv
    • while going down (dx < 0) negativ.
  • Turning Point:
    • the point around which the robot turns is fine while going up
    • while going down it lies way to far behind the chair, instead it should be on the front of the chair -> change by initializing dx_1 & dx_2 with a controlled offset or a constant alpha change once the alpha determination works
  • the wheelchair is usually quite unstable on the curved stairs this was tested -> Track adjustments might help out on this matter

(2015-xx-xx) - robustness test of stair recognition on stairs without solid backwalls

Test Criteria:
  • Recognize stairparameters and position on stairs
Results:

(2015-06-03) - dynamic test of the controller on stairs

Test Criteria:
  • Offset/Distance_Traveled [m]/[m] from straight line up/down during climbing straight stairs
  • Performance difference of going up and down
Results:
  • Wheelchair stays within +-1° straight on stairs
  • Offset was not measured during this first test

(2015-05-16) - static test of controller on stairs

Test Criteria:
  • has only been tested without driving on stairs due to safety measures
Results
  • Kp = 0.05 align the chair to -3° to 0° within 10 s.

(2015-05-16) - test multiple fmincon algorithms on their performance and accuracy

  • sqp is the fastest with enough accuracy

(2015-05-16) - test multiple FoV configs on their performance and accuracy

Test Criteria:
  • At least three edges need to be within the FoV to guarantee a correct identification of the stair parameters
    • This is dependant mostly on the stair depth
    • An initial ranging scan with a big FoV can be done to reduce the FoV to its optimal size
Results:
  • 150 seems to be the sweetspot
  • more testing and calculations still needed

(2015-05-16) - variable track and stair visualization are working

Test Criteria:
  • Stairs grow dependant on distance traveled
  • Tracks change in regards to their position at the acctual wheelchair