That's great! I never made such a thing because I was daunted by the task! Congratulations. In looking at the video, it seems to be well damped. Maybe overly damped as it takes a while to stop moving after getting a push. It looks as it there is a quick response which then damps alittle too fast. What happens with more P and/or a little less D?
-Tony
Odd 9V Power Drain
-
liam.zbasic
- Posts: 163
- Joined: 24 March 2008, 23:33 PM
- Location: Southern California (Blue)
In the case of the floor GBOT, Kp was approximately 10*Kd and Ki was nearly zero. Surprisingly, in the case of the balancing bot, the Integral gain (Ki) is 3 times higher than Proportional Gain (Kp), and Derivative gain is only a tenth of Proportional Gain (Kp). In other words, Ki = 3*Kp = 30*Kd.
I started GBOT90 with mostly Kp and some Kd, but the system was very jittery. Part of the issue was resolution. Filtering the IR signal in hardware & software was critical. Drift was less of an issue. Then I started dialing up Ki, and the jitter reduced, but the drift increased, especially in areas where my floor is not level.
So more Kp and some Kd results in less drift, but more jitter, and the Martini test would not have a happy ending.
The key is to add another IR sensor to reduce drift, or replace the IR approach with a single gyro. That hurts the wallet.
I started GBOT90 with mostly Kp and some Kd, but the system was very jittery. Part of the issue was resolution. Filtering the IR signal in hardware & software was critical. Drift was less of an issue. Then I started dialing up Ki, and the jitter reduced, but the drift increased, especially in areas where my floor is not level.
So more Kp and some Kd results in less drift, but more jitter, and the Martini test would not have a happy ending.
The key is to add another IR sensor to reduce drift, or replace the IR approach with a single gyro. That hurts the wallet.