Temporal-Spatial based Deep Anomaly Detection (TS-DAD) Model : A Steering Sensor Anomaly Detection in Unmanned Ground Vehicles

The operational efficiency of Unmanned Ground Vehicles (UGVs) profoundly relies on sensors for navigation and vehicular controls for functional operation. Sensors like brake, throttle, steering and speed sensors play a notable role in the mobility and agility of the UGV. Steering sensors among the other system sensors, are crucial for maintaining stable direction and maneuverability. Steering sensors are susceptible to major risk factors like internal faults, external disturbances, or cyber physical attacks which can cause serious risks to vehicle safety. The proposed Temporal-Spatial Deep Anomaly Detection (TS-DAD) model uses deep learning techniques with fuzzy logic to find anomalies in steering sensor data. The detection model captures both temporal and spatial characteristics of time-series data to improve detection accuracy. Experimental evaluation on real-time data demonstrates that the TS-DAD model achieves an accuracy of 98.2% which effectively identifies abnormal behavior in steering systems