Resilient Sensors for Measurement Assurance

Electronic sensors are transducers that measure environmental conditions, such as temperature, pressure, magnetic fields, acceleration, and electrical current. Autonomous systems use sensor information about surrounding conditions to make navigational and operational decisions in real-time using AI and Machine Learning algorithms. Sensors are subject to failure and accuracy drifts that have led to catastrophic results. A low-cost alternative to redundant sensors is presented that allows real-time detection of anomalous sensor data with low computational and electrical power. This alternative relies on the uniqueness within manufacturing tolerances of the sensor responses. A second low-cost uncalibrated sensor is used to monitor the performance of a higher-cost calibrated sensor. Differences between the sensors that exceed an initially measured threshold difference are flagged as an anomaly. Resilience can be added to the sensor system using arrays of sensors which allows the system to mask out faulty sensors. Larger arrays of sensors also allow for the generation of unique system identifiers which add security to data transmissions. A system architecture based on patented ideas and prototyped sensor systems is presented.