Biological cells are colloidal systems. In eukaryotic cells, cytoskeletal architectures contribute to diverse cellular activities, such as mitosis, cellular morphogenesis, and migration. Among these various activities, protein motors (kinesin, dynein, and myosin) of the order of a few tens of nanometers and protein filaments (microtubules and actin filaments) of the order of a few micrometers play important roles in the formation of dynamic architectures as the colloidal nature of living cells. However, the mechanism of their network formation and the relationship between the physical properties of protein motors/protein filaments and generated networks remain elusive. To reveal these relations, we have developed two model systems composed of a few types of purified cytoskeletal components. In one model system, purified protein motors, dyneins are immobilized on a glass surface. These surface-bound dyneins can drive microtubules on the surface in the presence of ATP. At high surface densities of microtubules driven by dyneins, microtubules were self-organized into the dynamic lattice of millimeter-scale vortices through their nematic interactions. Dynamics of the pattern depends upon the types of dyneins. The numerical simulations assuming nematic interactions among microtubules and memory of the curvature of their trajectories reproduced the experimental results. In the other model, the mixture of microtubules and tetrameric kinesin motors (Eg-5) generated three distinct 3-D network structures with millimeter scales, such as a static network, an active network characterized by self-rupture, and an aggregating network. The modification of motor properties led to the formation of many microtubule asters with the uniform size. To determine how the system property defines spatiotemporal dynamics, we constructed numerical coarse-grained models. Our models reproduced all of the experimentally-observed patterns and predicted the storage of the elastic energy in the cytoskeletal network and its abrupt release as mechanical work. Our experimental systems and numerical models bring new insights into not only cytoskeletal mechanics but also the development of self-regulating active materials.