Deep Learning Based Framework for Multicopter Sensor Data Imputation

The study presented in this paper introduces a novel approach for imputing missing sensor data in multicopters, which enables enhanced safety and reliability by leveraging the multitude of sensors on these aerial vehicles. The proposed approach is based on two deep learning techniques, namely Autoencoders (AE) and Long Short-Term Memory (LSTM) networks. The effectiveness of this approach is evaluated using flight test data from a 2.5 kg hexacopter, and three different scenarios of missing data are considered. To validate the performance of the proposed approach, it is compared against two commonly used imputation techniques: k-Nearest Neighbor (KNN) imputation and Random Forest imputation. The results indicate that the proposed approach outperforms both KNN and Random Forest in terms of the accuracy of imputation. The network has an error of less than 10% when processing signals with six missing sensor readings for a duration of 10 seconds. In contrast, KNN and Random Forest algorithms have an average error of 18% and 26%, respectively. Moreover, the trained model can handle missing data with varying degrees of sparsity, which makes it a more robust and flexible solution. The study also investigates the impact of using initial estimates provided by the Kalman Filter for training the deep learning models. It is observed that incorporating these estimates does not result in any improvement in the imputation accuracy. This suggests that the proposed approach is able to learn the underlying patterns in the data without the need for additional information from the Kalman Filter. Overall, the results of this study demonstrate the potential of deep learning techniques for imputing missing sensor data in multicopters. The proposed approach offers a more accurate and efficient solution than the traditional imputation techniques, and can handle varying degrees of data sparsity. The findings of this study have important implications for the design and operation of multicopters, and could result in enhanced performance and operational effectiveness of these aerial vehicles.