This paper presents the development of autonomous navigation for a self-balancing bicycle, considering its kinematic constraints. For this purpose, path planning is based on an evolutionary differential algorithm (LSHADE-COP), combined with the Non-Uniform Rational B-Spline (NURBS) curve representation. This approach provides a smooth, differentiable, and feasible path for the robot. In addition, a vector field-based guidance method is used to ensure that the robot follows the planned path. Simulation results show the importance of considering robot orientation and kinematic constraints from the path planning step, ensuring safe navigation. Besides that, initial experiments with the bicycle were conducted to validate the balancing and low-level control required for robot navigation.