Abstract: We decribe the dynamics modeling, attitude control, system simulation, and prototype of a rear-wheel-drive bicycle robot without a mechanically assisted structure. The robot maintains its balance solely by the handlebars for realizing its rectilinear steady running. Furthermore, the bicycle robot is a typical asymmetric under-actuated nonholonomic constraint system that cannot balance easily. First, we establish the simplified dynamic model via system dynamic principle analysis of Lagrange formulation. Through the linearization of the under-actuated subsystem of frame rolling angle and handlebar torque, combined with a fuzzy adaptive control algorithm, the controller is constructed with partial feedback linearization. The simulation and experiment results show that balance control during the simple rectilinear motion of the bicycle robot is realized effectively. This can lay a theoretical foundation for the balance control of bicycle robots and other under-actuated systems in the future.