Multivariate Satellite Telemetry Data Compression with Spatial Reduction Based on Linear Correlation-Guided Channel Selection and Receiver-Side Recovery

Satellite telemetry data compression is of great importance for preserving the integrity and effectiveness of telemetry information. As the number of telemetry parameters gradually increases, satellite states can be characterized from more comprehensive data dimensions. However, it is difficult to maintain long-term high-sampling-rate transmission for all channels constrained by telemetry bandwidth. This condition weakens the capability of telemetry data to capture satellite state variations in a sufficiently detailed manner. Therefore, it is still a challenge for selecting more representative telemetry channels and recovering the remaining dimensions. To address this issue, this paper proposes a data compression method based on linear correlation-guided channel selection and receiver-side Long Short-Term Memory Variational Autoencoder recovery, in which satellite telemetry data compression is achieved through spatial reduction. First, the Pearson correlation matrix of all channels is computed from telemetry data, and the structure among telemetry channels is analyzed based on a prior on pairwise linear relations. Second, a linear correlation-guided iterative channel selection strategy is developed under a fixed transmission budget. Candidate channels are scored by the residual correlation coverage matrix over the unselected channels, and the initially selected subset is further refined through local channel swapping to obtain a more globally representative transmitted subset. Finally, the transmitted channels are fed into the reconstruction model to recover the untransmitted channels, so as to ensure the transmission of telemetry information. Experiments are conducted on real multivariate satellite telemetry data under predefined transmission budgets. Results on 84-dimensional real operational satellite telemetry data show that the proposed method exhibits favorable data compression capabilities. Representative RMSE comparisons under different transmission budgets further showed favorable reconstruction performance, demonstrating the feasibility of the proposed method for telemetry data compression under limited communication resources.