Ambient radio frequency (RF) energy harvesting technology provides a hopeful solution to power the low-power electronic system. Nevertheless, the extremely low power density and randomness of ambient RF energy pose a significant obstacle to the application of ambient RF energy harvesting technology. Aiming at wireless transmission by ambient RF-powered sensors, this paper presents a cooperative communication technology by combining ambient backscattering and energy beamforming to significantly decrease the transmission energy consumption of each sensor. For wireless transmission via this cooperative communication technology, a novel optimization problem is formulated to maximize the long-term time-averaged sensed data size, which jointly takes into account sensing scheduling, transmission power allocation and the control of sensed data size. Moreover, we employ techniques of Lyapunov drift and Lyapunov optimization to propose an online optimal control algorithm, which achieves the optimal solution to the proposed problem without requiring a-priori knowledge of harvested energy and wireless channels. By exploiting mixed integer programming (MIP) and signomial geometric programming (SGP) that cause the proposed problem difficult to solve, a novel Sequential Convex Approximation algorithm is further introduced into the online optimal control algorithm. Simulation results indicate the proposed online algorithm offers a dramatic performance gain over alternative online algorithms.