In theory you can interpolate/accumulate at lower frequencies and obtain some 2mm semi-precision. Assuming that level does not changes rapidly over time. Or use different waveform, not fmcw. 70 GHz would be difficult even with ic. Easier with on-chip antenna. I personally do not like 70 GHz idea. Good engineering approach is to do all calculations on required power, etc from requirements. Do few design iterations. If you want fast results i suggest buying some sensor and try reimplement its dsp signal processing, it might be.not so easy. Then reimplement rf part. Also you can buy some rf Evaluation board for ic you mentioned. Actually it is years of learning and knowledge. I started to learn rf few years ago in my free time, and there always new questions arise. I do not know where to.start helping you. Maybe some pll and vco with receiving mixer and transceiver antenna evaluation board to play with it, give some feeling. Best results can be obtained with vna and spectrum analyzers, but they are very expensive.
Also you start with trying cheap sensors. Google for rfbeam and innosent.
- - - Updated - - -
Update:
you can start with buying:
1. Innosent or rfbeam sensor with modulation
2. Some cortex arm development board (stm32, tms320,lpc, etc.) With adc and dac.
Connect dev board to sensor, control sensor modulation using dac, get IF signal with adc, do some dsp processing, i guess it would be fft. After some time you will you will understand what is feasible, what part must be improved.