The Worldwide Journal of Robotics Analysis, Forward of Print.
This paper addresses the issue of decentralized, collaborative state estimation in robotic groups. Particularly, this paper considers issues the place particular person robots estimate related bodily portions, akin to one another’s place relative to themselves. The usage of pseudomeasurements is launched as a way of modeling such relationships between robots’ state estimates and is proven to be a tractable solution to strategy the decentralized state estimation drawback. Furthermore, this formulation simply results in a general-purpose observability check that concurrently accounts for measurements that robots gather from their very own sensors, in addition to the communication construction inside the workforce. Lastly, enter preintegration is proposed as a communication-efficient method of sharing odometry info between robots, and your entire principle is acceptable for each vector-space and Lie-group state definitions. To beat the necessity for speaking preintegrated covariance info, a deep autoencoder is proposed that reconstructs the covariance info from the inputs, therefore additional lowering the communication necessities. The proposed framework is evaluated on three totally different simulated issues, and one experiment involving three quadcopters.
发布者：Charles Champagne Cossette，转转请注明出处：https://robotalks.cn/decentralized-state-estimation-an-approach-using-pseudomeasurements-and-preintegration/
