The Bias frames are very short exposures made with no light reaching the sensor. With a DSLR, they are taken at the fastest possible shutter speed and lowest possible ISO, with the body, lens or telescope capped. There purpose is to remove the read noise that is present and roughly the same in every frame. You should take a large number of Bias frames (~50), and DSS will average them to create a master Bias frame, which it will subtract from every other frame used in the image processing. You don't need to take new Bias frames very often, just once or twice a year, because the read noise doesn't change much over the life of a DSLR. The Bias frames also contain the "Offset" signal, an elevated baseline added to every image by the camera electronics. With my Canon 350D, the Offset signal is 255 counts out of a maximum range of 4,096 counts (12-bits). So subtracting the Bias frame from every other shot removes the Offset signal as well as the read noise.
Dark frames are made at the same exposure time and ISO as the subject light frames, but with no light reaching the sensor. They should also be made at approximately the same temperature as the light frames. Normally they are interspersed in the imaging session, or taken at the end of the imaging session. Their purpose is to remove the thermal noise from the light frames. Normally you want between 10 and 20 dark frames during each imaging session, and if you use multiple exposure time or ISO values for the light frames during a session, you must acquire sets of dark frames to match the exposure and ISO of each set of light frames. The master Bias frame is subtracted from each of the dark frames before they are combined to create a master dark frame.
Flat frames are short exposures made through the imaging optics with the camera pointed at an evenly illuminated field. They record uneven field illumination due to vignetting caused by the optics and due to partial obstructions in the light path such as dust particles on the sensor, and they are use d to remove that uneven field illumination from the light frames. Flat frames can be taken by pointing the camera at the twilight sky, by covering the aperture with a smooth, translucent, light colored material such as a t-shirt, or by covering the aperture with a specially constructed "light box", and making 15 to 20 exposures in Av mode. They must be taken with the camera in exactly the same orientation relative to the telescope as used for the light frames, so they are usually taken at the beginning or end of each imaging session. ISO isn't critical but most people use the same ISO as used for the light frames. The master offset frame is subtracted from each flat frame, and then the flat frames are averaged to make a master flat frame. The master flat frame is divided into each of the light frames, after the master Bias and master Dark frames have been subtracted from the light frames.
Applying the Offset, Flat and Dark frames to the subject frames is called calibration. A few months ago I made a photo documentary of the calibration process on my PBase site, and you can find it here:
Click on the images to read the descriptions.