Mirror
/
ardupilot
mirror of https://github.com/ArduPilot/ardupilot
5
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

AP_AccelCal: fix bug preventing accel cal fit to run more than one iteration 

The check for fitness being smaller than the last value should use the
absolute value, to mean the fitness isn't changing since last iteration.
It's currently always quiting the function after the first iteration.

However for Gauss Newton we anyway want to run as many iterations as we
can, because there are ups and downs along the iteration so we might get
say best result where before and after iterations were bad. The lines
above takes care that sane and the best is selected.
This commit is contained in:
Siddharth Bharat Purohit 2016-11-18 13:01:00 +05:30 committed by Lucas De Marchi
parent 11229d2588
commit 451c2f146d
1 changed files with 0 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
libraries/AP_AccelCal/AccelCalibrator.cpp
View File

@ -334,8 +334,6 @@ void AccelCalibrator::run_fit(uint8_t max_iterations, float& fitness)
uint8_t num_iterations = 0; uint8_t num_iterations = 0;
while(num_iterations < max_iterations) { while(num_iterations < max_iterations) {
float last_fitness = fitness;
float JTJ[ACCEL_CAL_MAX_NUM_PARAMS*ACCEL_CAL_MAX_NUM_PARAMS] {}; float JTJ[ACCEL_CAL_MAX_NUM_PARAMS*ACCEL_CAL_MAX_NUM_PARAMS] {};
VectorP JTFI; VectorP JTFI;
@ -379,9 +377,6 @@ void AccelCalibrator::run_fit(uint8_t max_iterations, float& fitness)
} }
num_iterations++; num_iterations++;
if (fitness - last_fitness < 1.0e-9f) {
break;
}
} }
} }