Sensor Fault Detection and Diagnosis for Enhanced Safety of Autonomous Systems

Many modern systems such as autonomous vehicles and manufacturing systems operate under the control of a computer. All of these systems rely on sensors, which measure things like speed, temperature, and pressure. If one of these sensors fails, however, then the computer may take an incorrect action that can hurt people or damage property.

Therefore, it is extremely important to ensure all of the sensors that provide measurements for the computer are working correctly. This project aims to enhance the safety of these systems by developing methods for checking on whether the measurements provided by the sensors are correct and can be used safely by the computer.

Adaptive delayed left inversion (ADLI) constructs a causal, delayed left inverse of a dynamical system that represents the relationship between two sets of sensors, namely, input sensors, which are suspect, and output sensors, which are assumed to be healthy. Multiple combinations of sensors will be considered in order to determine whether the output sensors are indeed healthy.

Measurements from the healthy sensors are used to drive the delayed left inverse, whose output provides estimates of the expected measurements from the suspect input sensors. By comparing these estimates with the actual measurements, it is possible to detect and diagnose sensor faults. ADLI will be applied to discretized nonlinear kinematic differential equations that relate signals from multiple sensors, thus, providing the means for sensor fault detection and diagnosis.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.