openDog #13 Preview, for Patron only (Patreon)
Content
The whole video is long and detailed as usual, the summary is: I put new feet on which have space to have pressure sensors added, and I also fit an IMU and try to make the robot balance. Balancing on openDog is a bit harder than RobotX, even though it has more legs, because it's feet are round and not big and flat.
I'm using the IMU to affect X and Y translation, and also pitch and roll moves, to try to get it to stay level when it takes its feet off the ground. You'll remember that currently I have the motion filtering set on each actuator right before the positions are written to the ODrive, however this also smooths out the reaction from the IMU/PIDs making it less effective.
As I reduce motion smoothing it gets better at balancing, however as I speed up the leg motions the un-smoothed motions cause issues with both inertia and the speed that its feet hit the ground. I sort of got a happy medium at the end, but it's not ideal. So future actions are:
-Upgrade the 8-bit Arduinos to faster 32-bit micros, probably Teensys TBC.
-Restructure the code so that the motion smoothing filter happens on the initial requested moves, but the stability correction from the IMU/PIDs happen after that. This means that the motion is smoothed, but the reaction to balance is quick.
-Investigate the 'interpolation engine' I mentioned in openDog #12, or at least proportional speed control for each actuator so that all the actuators stop moving at the same time for any foot XYZ position. ODrive also has 'trajectory planning' in firmware development which may be similar, although I don't know if it'll work in three dimensions across multiple ODrives.
-Reduce the amount of things strung together on serial lines, or at least increase the baud rate and cycle time on each micro. The ODrive can receive commands 4000 times per second before there's a buffer overflow, so there's lots of scope to tighten things up here.
-Decide if foot pressure sensors can help with at this stage (they definitely can, but I have to decide what to do with the data).
The next video in the series is probably looking at Teensy and another ODrive which I've purchased to see what the most 'minimal microcontroller way' of doing things is. CANbus and actuator end stops/homing switches are also in development in ODrive firmware, so it's possible I can reduce everything to one microcontroller rather than four.
Files
Previews only