Attitude determination with low-cost GPS/ INS

Low-cost GNSS receivers with patch antennas track the carrier phases of the GNSS signals with millimeterto centimeter-level accuracy. However, code multipath of several tens of metres, frequent half cycle slips, and receiver clock offsets in the order of milliseconds make reliable kinematic integer ambiguity resolution still challenging. Low-cost inertial sensors are robust against GNSS signal shadowing and benefit from a higher measurement rate but show large integration drifts due to biases in the gyroscope and acceleration measurements and, require an initialization with another sensor. In this paper, we couple the information of both sensors for attitude determination and propose a search algorithm for cycle slip detection and correction. The search algorithm uses double difference carrier phase, acceleration and gyroscope measurements, and also considers a priori information on the baseline length. The cycle slip corrections are determined such that the optimum trade-off between minimizing the squared measurement residuals and minimizing the squared baseline length residuals is found. We tested the method during various car drives. Measurement results show that the method reliably corrects all cycle slips in environments with both high multipath and high receiver dynamics, and enables a heading determination with an accuracy of 0.5/ baseline length [m]. INTRODUCTION Attitude determination is widely performed with inertial sensors and geodetic GNSS receivers in maritime and aerospace applications. The carrier phase can be tracked with millimeter-level accuracy and, therefore, is very attractive for precise attitude determination. Double differences are formed to eliminate receiver and satellite clock offsets and biases as well as atmospheric errors. This leaves the double difference integer ambiguities and baseline coordinates as unknowns. Teunissen developed the Least Squares Ambiguity Decorrelation Adjustment (LAMBDA) method in [1] to solve the integer least-squares problem. For attitude determination, the success rate of ambiguity resolution can be further increased if the baseline length a priori information is included in the search process. In [2] and [3], Teunissen developed a constrained LAMBDA method to solve the constrained integer least-squares problem. Low-cost GNSS receivers can also track the carrier phase with millimeterto centimeter-level accuracy. Consequently, they are very attractive for automotive applications (e.g. heading determination of cars, Fig. 1).