Wikipedia gives a very good explanation of this, on the page for the suprachiasmatic nucleus.
For example, in the fruitfly Drosophila, the cellular circadian rhythm
in neurons is controlled by two interlocked feedback loops.
In the first loop, the bHLH transcription factors clock (CLK) and
cycle (CYC) drive the transcription of their own repressors period
(PER) and timeless (TIM). PER and TIM proteins then accumulate in the
cytoplasm, translocate into the nucleus at night, and turn off their
own transcription, thereby setting up a 24-hour oscillation of
transcription and translation. In the second loop, the transcription
factors vrille (VRI) and Pdp1 are initiated by CLK/CYC. PDP1 acts
positively on CLK transcription and negatively on VRI. These genes
encode various transcription factors that trigger expression of other
proteins. The products of clock and cycle, called CLK and CYC, belong
to the PAS-containing subfamily of the basic helix-loop-helix (bHLH)
family of transcription factors, and form a heterodimer. This
heterodimer (CLK-CYC) initiates the transcription of PER and TIM,
whose protein products dimerize and then inhibit their own expression
by disrupting CLK-CYC-mediated transcription. This negative feedback
mechanism gives a 24-hour rhythm in the expression of the clock genes.
Many genes are suspected to be linked to circadian control by "E-box
elements" in their promoters, as CLK-CYC and its homologs bind to
The 24-hr rhythm could be reset by light via the protein cryptochrome
(CRY), which is involved in the circadian photoreception in
Drosophila. CRY associates with TIM in a light-dependent manner that
leads to the destruction of TIM. Without the presence of TIM for
stabilization, PER is eventually destroyed during the day. As a
result, the repression of CLK-CYC is reduced and the whole cycle
reinitiates again. (http://en.wikipedia.org/wiki/Suprachiasmatic_nucleus#Fruitfly)
The suprachiasmatic nucleus (SCN) is a tiny part of our brain residing in the center. It maintains a biological clock through a gene expression cycle in the individual neurons. The mechanism for humans is very similar to the mechanism for fruit flies, as explained above, but to rehash a bit...
In the fruit fly model there are five players: CLK, CYC, PER, TIM, CRY.
CLK and CYC are transcription factors for PER and TIM, and bind to their promoters in order to activate the expression of PER and TIM.
When the expression level of PER and TIM gets very high, the PER and TIM proteins return to the nucleus, and inhibit their own transcription factors (CLK and CYC) through molecular interactions.
The CRY protein is light-sensitive, and so daylight (or artificial light) will cause CRY to destroy the TIM proteins. Without TIM, we no longer get the PER-TIM dimers that inhibit the CLK-CYC transcription factors.
In humans and other mammals it's the same concept, but with slightly different players (e.g., homologous genes).
What's important here is that we have a biological clock on the cellular level. We haven't gone into how this clock is used yet... But, it's my understanding that the SCN uses the information to tell the pineal gland when to produce melatonin.
The melatonin causes us to be drowsy, lowers our body temperatures, and causes us to fall asleep.
You can see another cycle at work here - our body temperature, which rises with wakefulness, and declines with sleep.
This gives a more general answer to how circadian rhythms are used to influence our sleep cycle. Your question was how the genes involved in circadian rhythms regulate each other. The answer is the transcription-translation negative feedback loop described above.
Hope this helps...