NRI: Receding Horizon Integrity-A New Navigation Safety Methodology for Co-Robotic Passenger Vehicles

Illinois Institute Of Technology, Chicago IL

Investigators

Abstract

The objective of this research is to ensure the integrity of vehicle position, heading, and velocity estimates that are used by self-driving cars as the basis for life-critical decisions such as the initiation and execution of hazard-avoidance maneuvers. Integrity, which is a measure of trust in a sensor's information, has been successfully implemented in commercial aircraft to guarantee the safety of maneuvers such as landing. This project addresses several obstacles in translating integrity from aviation applications to self-driving cars, including integrating the disparate sensor types used by ground vehicles; meeting the stringent demands of routine autonomous driving; accounting for the number, proximity, and high relative velocity of other vehicles on the road; and evaluating multiple, distinct, and mutually exclusive courses of action in a timely manner. Project subtasks include characterization of integrity for representative sensors, construction of appropriate models for uncertainty propagation, and experimental validation of the resulting integrity framework. The project will advance the larger research effort to realize the potential of self-driving cars for relieving congestion, reducing emissions, and saving lives. The work includes public outreach efforts on autonomous navigation for self-driving cars, which will build upon an ongoing relationship with Chicago's Museum of Science and Industry, including a hands-on demonstration during National Robotics Week to illustrate how safety can be ensured despite uncertainties related to sensor readings, vehicle dynamics, and the driving environment. Specifically, this research will provide new experimental and analytical methods to quantify and prove self-driving car safety. The results of this work will create a high-level, sensor-independent, quantifiable metric that can be used to compare, evaluate, and certify safety across self-driving car manufacturers. Knowledge will be advanced in several previously-unexplored areas, including first-ever demonstrations of: 1) high-integrity sensor measurement error and fault models for non-GPS sensors, 2) analytical methods to quantify the safety risk of feature extraction and data association algorithms required in lidar, radar, and camera-based localization, 3) multi-sensor pose estimators and integrity monitors designed to evaluate the impact of undetected sensor faults on safety risk, and 4) rigorously derived and experimentally validated integrity risk prediction methods in dynamic environments.