Robotic sensor systems play a vital role in the intelligent working of service robots, and service robots are important components of smart cities. Robotic sensors sense information from themselves and their environment, and fuse pieces of information for use in perception, decision, planning, and control. In this paper, we investigate the application of robotic sensor systems in 3D biped walking. By using sensory feedback control, offline gait planning and online gait modification schemes have been proposed. The gait planner plans the frontal and lateral reference trajectory of a biped robot. Moreover, the kinetic constraint, zero moment point (ZMP) stability criterion, and unilateral constraint are taken into account, simultaneously. The gait modification consists of the torso posture controller, the ZMP compensator, and the joint angle controller. It is utilized to reduce the robot joint error through posture control and ZMP compensation. Additionally, the online sensor parameter optimization based on reinforcement learning (RL) is proposed to improve the stable margin of walking. Finally, simulations are presented to illustrate the effectiveness of the proposed method.