4 results about "Time step" patented technology

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a prediction characterizing an environment. In one aspect, a method includes obtaining a respective observation characterizing a state of an environment for each time step in a sequence of multiple time steps, comprising, for each time step after a first time step in the sequence of time steps: processing a network input that comprises observations obtained for one or mor preceding time steps to generate a plurality of acquisition decisions; obtaining an observation for the time step, wherein the observation includes data corresponding to modalities that are selected for acquisition at the time step, does not include data corresponding to modalities that are not selected for acquisition at the time step; and processing a model input that includes the observation for each time step in the sequence of time steps to generate the prediction.

This application discloses a method and related apparatus for predicting channel time series at future times. The method includes: firstly, acquiring the channel time series at M times to obtain X(t) m )={x n (t m Let X(t) = |m=1,2,…,M} (n=1,2,…,N), where N is the number of channel time series within a given time moment, and M and N are both positive integers. Next, based on the stated X(t)... m )={x n (t m Construct a mapping Ψ for each m = 1, 2, ..., M (n = 1, 2, ..., N). p , among which, Ψ p (X(t m ))=x n (t m+p‑1 (p=2,3,…,L), where L-1 is the number of time steps to be predicted, and X(t) is used as the time step to be predicted. m (m=1,2,…,M-p+1) as input, with x n (t m+p‑1 Using (m=1,2,…,M-p+1) as the output, we perform a Gaussian process regression (GPR) fitting to obtain Ψ. p (p = 2, 3, ..., L). Finally, X(t) m (m = M - p + 2, M - p + 3, ..., M) respectively input the Ψ p (p = 2, 3, ..., L), we get x n (t m+p‑1 This application generates predicted values ​​for (n = 1, 2, ..., N; m = 1, 2, ..., M; p = 2, 3, ..., L), thus achieving multi-step prediction of channel time series. Compared with existing single-step prediction, this application does not accumulate or amplify errors during the prediction process, resulting in more accurate prediction results.

A method executable by one or more processors includes: obtaining a measurement of a first variable at a current time step; using a first model to estimate a second variable at the current time step based on the measurement of the first variable; generating a control decision for a subsequent time step based on the measurement of the first variable and the estimate of the second variable using a reinforcement learning model; using a second model to predict a predicted value of the first variable for a subsequent time step based on the measurement of the first variable at the current time step; adjusting the control decision for the subsequent time steps based on constraints and future values ​​of the first variable; and controlling an actuator based on the control decision.