We start where the last post ended, the closure of the Panama gap around 2 million years ago which led to a loss in transport of warm water to the north pole. Erosion of the newly-formed mountain ranges such as the Himalayas slowly reduced the CO2 in the atmosphere and the earth slowly grew colder with ice building at the poles 
The causes of ice ages are not completely understood . A strong theory is that slight changes in solar radiation being received due to orbital changes can then amplified through a number of positive feedback processes. The reflection of sunlight back from ice and snow reduces the net heating of the sun, and the formation of large ice sheet reduces the flows of oceans to and from the ice sheets leading to further increase in ice formation. Analysis of trapped air bubbles shows that greenhouse gases drop by a third (CO2) and a half (methane) during glaciation. There is some dispute on whether CO2 levels lead or lag the temperature changes, though there seems to be some consensus that in ice ages it lags – the forcing is the solar radiation which drives temperature, and the amount of CO2 in the atmosphere follows the change in temperature.
Building an ice age can take 80,000 years, but getting rid of ice can occur much faster – only 4,000 years. There are a number of reasons for this speed of deglaciation but the major reason would seem to be the rise in sea levels undercutting the ice sheet and causing more melting leading to a rapid melting of the ice sheet. During this deglaciation the water level increases massively – lowest to current is about 120m , so at some point the Mediterranean was a lake, then a salt basin, then flooded again.
The frequency of ice ages has been much analysed. They are clearly periodic but not so regular as to make explanation easy. The most obvious source of a periodic forcing variable is oscillations either in the earth’s orbit round the sun or from the sun itself. There are three separate cycles that interact: changes in eccentricity (the ratio of the length of the earth’s orbit to the width, the orbit being an ellipse not a circle); changes in the tilt of the earth’s axis of rotation, and precession of the tilt. These periodic oscillations known collectively as Milankovitch cycles interact to make ice ages more or less likely. The periods of oscillation of these aspects of the earth’s rotation are 96,000 years, 41,000 years and 27,000 years respectively. The combined effects are sufficiently vague in frequency and amplitude to be capable of continued re-interpretation . Furthermore if all else fails then meteors or super-volcanoes can always be invoked to explain a sudden cooling 
I’ve attached a chart that illustrates the temperature and CO2 . We are in the thick blue blob on the RHS. Possibly the most interesting part of the chart is the massive increase in CO2 on the RHS to a level not seen for the duration of this chart (400,000 years) which I'll return to shortly.
As we approach the Holocene (the name given to our current inter-glacial period starting nearly 12,000 years ago) we get a feel for the way temperature climate changes work – a forcing variable, then positive feedback mechanisms such as changes in ice and CO2 that drive the change until it reaches a new equilibrium. Then a switch back in the forcing variable until a new equilibrium is reached.
 http://www.bbc.co.uk/science/earth/water_and_ice/ice_age https://wattsupwiththat.com/2012/04/11/does-co2-correlate-with-temperature-history-a-look-at-multiple-timescales-in-the-context-of-the-shakun-et-al-paper/