How often you should calibrate your pipettes depends on the tolerances of your application. For some applications, like quantitative PCR setup, one may care a great deal; for general lab work, one can probably be less particular. You can get a sense of how accurate your pipettes are by pipetting DI water onto the platform of a sensitive (and calibrated!) balance. Comparing the random and systemic error you get while doing that over several trials to the specifications published by the manufacturer will give you a good idea of whether pipette calibration will help and help give you a sense of how much you care. Note also that the error may vary with the volume; you should test error over the entire volume range you intend to use the pipette for.
How to clean and calibrate pipettes is specific to each design. Some pipettes, like Gilson's, are not intended to be calibrated by the user. Others are. The pipette user guide published by the manufacturer should tell you.
For some demanding applications, even well-calibrated and well-handled micropipettes may not be sufficiently accurate. In particular, I've made standard curves by serial 10-fold dilution for quantitative PCR both a) relying on my micropipettors to be correct and b) trusting pipettes only to get me into the right ballpark, and intentionally undershooting and correcting volumes with sequentially smaller pipettes until the reading on the balance is correct, if that's at all clear. In both cases I was using 900 ul of diluent and 100 ul of the previous dilution, which I feel are comfortably large volumes to handle. I found that the latter curve gave me noticeably better regressions.