Filtering and fusion of consensus-based multi-agent systems with imperfect constraints

摘要

In state estimation (filtering) applications, state vector of a target always includes some a priori constraints, and utilizing such constraints will help obtain a more precise estimation. While in some cases, a significant issue lies in that when such constraints are not perfect, how we exploit them and improve the performance of estimation. Information contains “imperfection” sometimes because our knowledge of the real world is not absolutely right. This essay proposes a deep research on multi-agent systems with imperfect constraints and highly artificial intelligence: firstly, by using the projection method, results of traditional Kalman Filtering are projected onto the constrain subspace; secondly, based on consensus, this essay has thoroughly introduced the local information exchange and fusion strategy; and thirdly, simulations on target tracking have proved the feasibility and precision of the algorithms mentioned within this essay.