Privacy-Preserving Average Consensus via Finite Time-Varying Transformation

The privacy-preserving average consensus problem in multi-agent systems means that all agents in the system can reach an agreement on the average of their initial values, but guarantees that their initial states are private. In this paper, a new approach is proposed to solve this problem, furthermore, it can solve both asymptotic time consensus problem and prescribed-time consensus problem. The core idea of this approach is to construct functions which can mask the real states of agents in finite time. That is, within a certain period of time, what agents transmit is no longer their real state values, but the values acting on the designed functions. These designed functions are time-varying, local (i.e., determined independently by each agent), and converging to the true states in finite time. It is proved that the correctness (accurate calculation of global average value) and privacy (the initial values of agents are not speculated by other agents) can be guaranteed under our method. Finally, some numerical simulations are given to verify the effectiveness of our approach.