COLD-WORLDS aims to use gravitational microlensing to explore a unique niche, cold planets down to Earth mass orbiting around any kind of star, at any distance towards the Galactic center, rogue planets and moons orbiting exoplanets (exomoons). These are in very different environments from most known exoplanets, allowing key tests of planet formation theory. Indeed, the maximum sensitivity is for planets at the snow line, close their formation location. To date, 55 microlensing planets have been published and these results challenge theories of planet formation. The core accretion population synthesis predictions by Ida’s and Bern’s groups are quite similar and both under-predict the number of observed cold planets at a mass ratio of q =2E-4) by a factor of ~25. It might be due to the run-away gas accretion phase of planet formation, which is a basic feature of the core accretion theory. Alternatively, it could be that there is some host star mass dependence of this run-away gas accretion gap that smooths out this feature when plotted as a function of mass ratio. So, it is important to accurately determine the individual masses for the planets and host stars. Microlensing provides precise mass-ratio and projected separations in units of the Einstein ring radius. In order to obtain the physical parameters (mass, distance, orbital separation) of the system, it is necessary to combine the result of light curve modeling with lens mass-distance relations and/or perform a Bayesian analysis with a galactic model. Often, physical parameters are determined to 30-50 %, or even worse. However, we have shown that a tight constraint can be obtained on the lens mass-distance, thanks to detection or upper limits on its luminosity using high angular resolution observations with 8m class telescopes or HST. The pioneering work by our team shows that we can derive physical parameters on known systems to 10 % or better with Keck adaptive optics for instance. In the uncertainty budget, we would then be dominated by extinction correction, distance to the source, calibrating luminosity function of main sequence stars and our understanding of the galactic structure. COLD-WORLDS will use infrared wide field imagers (public surveys from VISTA, UKIRT, and dedicated observations) and operate adaptive optics on 10m class telescopes. We obtained data already on 30 systems (Keck, VLT, SUBARU, HST) and we have 10 nights approved as Key Strategic Mission Support to WFIRST with Keck for the years 2018-2019. By 2019, we will have observations of the host stars of ~100+ systems with cold planets. We will use Gaia DR2 to measure the source distances, revisit the galactic disk, bar, bulge. Combining Gaia with the multiband observations, we will revisit our model of extinction. It will also give the kinematics on the line of sights to the Bulge and will allow to revise our Bayesian modeling of microlensing plane. We will then perform demographics of the Disk and Bulge cold planet populations and address the following science objectives. 1/ What is the mass distribution of cold planets down to ~1 Earth mass at the snow line, where most planets are formed? 2/ What is the spatial distribution and abundance of cold planets towards the centre of our Galaxy? 3/ How to routinely achieve better than 10% accuracy in microlensing planet mass determinations with the next generation of satellites (Euclid and WFIRST)? 4/ Producing a near-infrared (JHK) photometric archive from ESO dedicated surveys of the Galactic bulge, a lasting resource for broad areas of stellar and galactic astronomy. Our highly dedicated team covers all the range of needed expertise. It is also using public data and has all the needed allocated telescope time. It is a risk free, high impact project, giving roots to the Euclid and WFIRST microlensing surveys, while providing important results and useful legacy products.