A Robust Approach to Multiple Sensor Based Navigation for an Aerial Robot

This paper describes a multiple sensor fusion approach in which a sensor based navigation scheme needs to fuse a stochastic aerial robot position estimate from an extended Kalman filter (EKF) with a deterministic aerial robot position estimate from an interval analysis (IA) algorithm. The aerial robot is equipped with inertial sensors (INS) and ultrasonic sensors. An EKF is used to estimate the aerial robots position using the inertia! sensors. When landmarks are present, the ultrasonic sensor measurements are processed using an IA algorithm to get an interval aerial robot position estimate. In order to obtain a better estimate for the aerial robot position both deterministic and stochastic estimates need to be used via a data fusion approach. Thus there is a need to study how to fuse the aerial robot position estimate having a Gaussian distribution (from EKF) with a aerial robot position estimate that has a uniform distribution (from IA). This is accomplished here by using the Box-Muller transform to transform the interval aerial robot position estimate having a uniform distribution to a real number aerial robot position with a Gaussian distribution and giving that as a measurement to the EKF to obtain a fused estimate of the aerial robot position