There exist some ways to create a walking engine that balances. My goal is to combine two popular approaches, using a ZMP approach to do some balancing of a (predefined) movement and using an inverted pendulum approach to generate an intelligent foot step pattern. The idea is, that both ways have it's advantages, intelligent foot step planning makes it easily possible to compensate some (!) disturbing influences and the ZMP approach to balance the movement (i.e. inverse kinematic movement) and especially compensate the problems, where the other approach is inappropriate or can not operate, due to practical (constraint) limitations.

It is planned to work at first on each component separately and then, finally, merge these two approaches together, especially to use "desires" about the future ZMP-position (given by the inv. pendulum) to improve the basic ZMP-based balancing.

Hi, today I started developing a rudimentary balancing engine, that is needed, because we have a deadline for a first balancing engine. This means that we use a simpler approach and work on more advanced stuff later. Actually it seems that the first balancing engine consists just of a PID controller. But more about that later ... — schauert, 07.03.2007 20:25

Today the code of the first version, using a PID controller to control the robot inclination got finished. It will be interesting to see the first tests, if the sensors are built in and running ... Another thing that will be interesting to test, if a PID regulating the angular rate is better (that's what I would expect). — schauert, 08.03.2007 19:53

Ok, the German Open 2007 are over, I have to admit that the balancing engine had some problems, as a result of a very nasty (and frustrating) bug.

The last days/weeks I was porting some code - math, balancing engine, sensor data processor and various required code fragments - to our new framework and began to acquaint myself with it. This has nothing to do with balancing ...

... so let's get to the real information. I did some things about balancing, too. Some time ago I sorted out some papers/methods that seemed very promising to me. Now I got through some papers again and marked some methods as impossible to realize with our sensor arrangements. Some other time was invested into some math stuff - implementing some own stuff (e.g. matrix inversion, ...) as well as trying out some math libraries (e.g. BLAS, LAPACK). — schauert, 23.05.2007 23:30