Spoofing interference detection method for satellite-based train positioning based on CS-BP
With the continuous development need of train operation control system with the higher autonomy level of the train-borne sub-system and intelligent capability,the design and implementation of the novel train operation control systems require more autonomy and credibility capabilities of train speed and localization determination.Considering the advantages of the Global Navigation Satellite System(GNSS)with the wide coverage,high positioning accuracy level and the all-weather capability,satellite-based positioning has become an important development direction for the implementation of autonomous train positioning in novel train control systems.Due to the significant vulnerability of the satellite navigation systems,the interference to GNSS from the operational environment along the railway line will pose a serious threat to the performance of train positioning and state perception.Therefore,accurate and timely detection and identification of the existence of the GNSS interference has been a key issue to realize effective protection against the interference and ensure the safe train operation.The GNSS spoofing attacker may broadcast fake satellite signal information through signal retransmission or customized signal generation,which can deceive the GNSS receiver to calculate the wrong positioning result.Thus,the spoofing interference is more complex,convert and hazardous,which has a significant impact on specific GNSS devices.In the paper,a spoofing detection method for GNSS-based train positioning was proposed.Firstly,the features of spoofing interference to satellite-based train positioning are analyzed.Secondly,the typical feature quantity was established by extracting specific GNSS observations that are sensitive to the interference attack.Then,the overall spoofing detection scheme was designed,including sample set construction,off-line model training,on-line detection and recognition.Under this framework,a GNSS spoofing detection model construction method was proposed based on the BP neural network that is improved by the Cuckoo Search(CS)algorithm.Finally,a test environment for spoof injection to satellite-based train positioning was built,and a variety of observation information and characteristics of satellite positioning were fully used to identify the existence and characteristics of specific spoofing interference.The performance of the proposed detection method was examined and compared through spoofing injection tests.The results showed that the Cuckoo search strategy can expand the search scope,jump out of the local optimal and achieve fast iterative convergence.The convergence optimization performance by CS is less affected by the parameter conditions.The CS-enhanced BP neural network modeling solution can achieve the higher modeling performance than the conventional map matching residual threshold-based strategy and representative Machine Learning(ML)based methods.The detection accuracy of the proposed solution under different spoofing attacks is more stable,with better performance in time sensitivity,accuracy and F1 score.The proposed solution is capable of providing great support to the spoofing detection and interference protection in satellite-based train positioning and the location-enabled applications.It can provide an effective way to reliably utilize the satellite-based positioning and perception technology in novel railway train control systems.It can make it possible of reducing the trackside equipment like balises and track circuits,effectively reducing the cost of construction and maintenance,and improving the autonomy,flexibility and safety of the train-borne sub-system under complicated train positioning observation environments.
万方数据
维普
