The results are presented here of multi-wavelength observations centred on two fields of the Leiden-Berkeley Deep Survey, which form the basis for a study of the population and cosmic evolution of the high redshift, low power, Fanaroff & Riley class I (FRI) radio galaxies. These fields, Hercules.1 and Lynx.2, contain a complete sample of 81 radio sources with S_1.4GHz > 0.5 mJy within 0.6 square degrees. Wide-field, ~1.5 arcsec resolution, radio observations, along with near infra-red and optical imaging, and some multi-object spectroscopy, are used to select the best high-redshift FRI candidates, giving 37 in total. Currently, the host galaxy identification fraction is 86% with 11 sources remaining unidentified at a level of r > 25.2 (Hercules; 4 sources) or r > 24.4 (Lynx; 7 sources) and K > 20. Spectroscopic redshifts are determined for 49% of the sample and photometric redshift estimates are presented for the sample sources without spectra or previously published results. 95% of the 37 best FRI high-redshift candidate sources were then observed using sub-arcsecond radio resolution, with the aim of detecting extended emission with respect to compact core features - vital for unambiguous morphological classification. The nature of the radio observations meant that 10 extra sample sources could also be included in the data. Lower resolution data were also taken for the Lynx.2 field sources to provide a comparison with the 1.5 arcsec data. The classification of the entire radio sample is done in two stages. Sources which showed clear extension are classified by morphology alone, whereas sources with no obvious or weak extension were classified using a combination of morphology and flux density loss in the higher resolution data indicative of resolved out extended emission. Five groups are used for this - 'Certain FRIs', 'Likely FRIs', 'Possible FRIs', 'Unclassifiable sources' (for those not included in the higher resolution observations) and 'Not FRIs'. The final group numbers are 8, 10, 24, 33 and 6 for groups 1-5 respectively. The space densities of the maximum, probable and minimum FRIs are then calculated and compared to two previously published measurements of the local value, and with the behaviour of the strongest FRII sources. The results for all three groups show density enhancements of factors of 5-9 at z ~ 1.0 which implies cosmic evolution of the FRI population; this enhancement is also in very good agreement with that predicted by previous models. The behaviour of the FRI/FRII dividing luminosity, as a function of host galaxy absolute magnitude, at the different cosmic epochs of the sample, and for two different star formation histories, is also investigated. A shift to brighter absolute magnitudes is found to be inconsistent with the data but this conclusion is weak due to the lack of knowledge of the host galaxy stellar populations, and the small number of sources in the sample.