No abstract

No abstract

No abstract

The advancement of in-car navigation systems has dramatically improved driving experiences. However, ensuring the safety of these systems remains a critical concern. Federated learning provides a new solution for cooperative learning between non-mutually trusted entities. Through the mode of local training and central aggregation, the local data privacy of each entity is protected while training the global model. To achieve this, a federated learning method for deep learning that preserves privacy is developed by integrating differential privacy with secure multi-party computing. In this scheme, vehicles add perturbations to the local models obtained by local training and secretly share them with multiple central servers. The scheme protects the local information uploaded by users from being stolen and prevents the adversary from malicious inference from globally shared information such as the aggregation model. Additionally, the scheme enables users dropping out and implements a variety of aggregating methods. The aforementioned system may also easily be expanded to decentralized scenarios for real-world applications devoid of a trustworthy center. The experimental findings show that, in order to protect sensitive data obtained from in-car navigation systems during learning, the suggested strategy heavily emphasizes privacy protection. Simultaneously, the high accuracy achieved through the proposed federated learning scheme significantly enhances in-car navigation safety systems' detection and control capabilities. It enables precise and reliable event detection, differentiation of abnormal situations, and reduces false alarms, improving overall safety, user trust, and system performance.

In this study, we introduce how to detect broken wires in steel rope based on wavelet transform and virtual instrument technology. By means of the powerful data analysis function of virtual instrument and wavelet transform, the singularity of wires can be found and it could help to improve ability of locating broken wires and determining breakage grade.