Dislocation, a typical line defect in many solid materials, is considered as one major carrier of the plastic flow in materials under deformation. The mobility of it intrinsically determines how fast and how easy a material can be plastically deformed. Therefore, in order to predict the performance of engineering materials in real applications (finite-stress loading, non-zero temperature, and a combination of them), one must understand the stress- and temperature-dependence of the dislocation dynamics from the bottom up. At a fundamental level, the motion of dislocations and their interaction with each other in materials under deformation at finite temperature are inherently multiscale: it involves the dislocation nucleation at the atomic scale, kink (a deviation from a straight dislocation line) activation at the nanoscale, the long dislocation line migration, and their long-range interaction with each other at the microscale. Up to date, a simulation tool that can simultaneously resolve the μm-long dislocation migration and the atom-sized kink activities along the long dislocation line does not exist yet. To meet this need, the goal of this dissertation research is to develop a coarse-grained (CG) method for modeling the dislocation dynamics in plastically deformed materials at finite temperature. One unique feature of the newly developed CG is that it will not only describe the dynamics of a μm-long dislocation line, but also retain the atomistic nature of dislocation core reconfiguration, kink activation and annihilation along the dislocation line itself. Fundamental to this CG method is the finite element implementation of an atomistic field formulation that unifies the atomistic and continuum description of materials through an Irving-Kirkwood procedure in statistical mechanics. To explicitly capture the thermally activated kink activities along the long dislocation line in high-Peierls-barrier materials under deformation at finite temperature, a phonon density states-based temperature controlling algorithm is developed and implemented into the CG models. With the interatomic force field being the only input and at a fraction of the cost of full molecular dynamics, the capability of the newly developed CG method is demonstrated through a series of simulations. Taking bcc tungsten (W) and iron (Fe) as model materials, we quantify the kink-controlled dynamics of ½ <111> screw dislocations with their lengths ranging from 30nm up to 3μm through finite temperature coarse-grained (FT-CG) simulations at only a fraction of the cost of quantum and atomistic calculations. Our major findings are: (i) the kink activation enthalpy, ΔH, is a decrease function of the stress, which agrees well with that from density functional theory (DFT) calculations. ΔH for kinks along the long dislocation lines is lower than that on short ones. Such a length dependence of ΔH appears in W but not in Fe; (ii) a four-stage atomistic process, i.e., multi-kink pair formation, cross-kink, self-pinning, and then the debris production, only occurs on the μm-long dislocations, but does not appear on the nm-long dislocations; (iii) the commonly assumed linearity in the length dependence of a dislocation’s mobility vanishes, especially for the dislocations with intermediate lengths (hundreds of nanometers) in W subjected to a high stress at high temperature, where the kink-pair mean free path and the dislocation length become comparable; (iv) a generally applicable mobility law for high-Peierls-stress dislocations needs to incorporate not only temperature-, stress-, but also the line length independence, linearity, and nonlinearity all in a closed form. We consolidate these findings into a series of maps for correlating stress, temperature, and the length of a screw dislocation segment with its mobility in one diagram. These maps all contain two strikingly different regimes according to two distinct modes. The boundary between these two regimes largely shifts when the dislocation is micrometer long. We consider this work as a first attempt to address the full complexity of the kink-controlled dislocation dynamics from the atomistic to the mesoscale. It may provide a bridge to connect atomistic with the higher length scale models for understanding the macroscale plasticity not only in Fe and W, but also in semiconductors, ionic crystals, high entropy alloys, and many more, whose deformation physics at low temperatures are all largely dictated by μm-long kinked dislocations.