Sensor Fusion in Autonomous Navigation for Asteroid Observation Missions

Abstract

Spacecraft navigation and control is difficult in deep space operations. Especially around asteroids, the irregular gravity field increases the difficulty of estimating the spacecraft trajectory. Autonomous navigation can increase the safety and accuracy for orbit proximity operations. Furthermore, it eliminates the need for continuous communication with the spacecraft. For the implementation of autonomous navigation in deep space, the onboard guidance navigation & control (GNC) should be able to accurately estimate the attitude and relative position of the spacecraft. By using sensor fusion, information from individual sensors can be combined to increase certainty and accuracy of the state estimation. A sensor fusion model is proposed, comprising an inertial measurement unit (IMU), star tracker and light detection and ranging (LiDAR) as navigation sensors. The aim of this research is to investigate the feasibility and performance applying sensor fusion for the spacecraft state estimation.

A navigation filter is applied to a benchmark scenario that orbits an asteroid at 50 km. In this scenario, asteroid 433 Eros has been selected for its unique shape, which has been mapped during the Near Earth Asteroid Rendezvous (NEAR) mission. A simulation is performed to approximate the dynamics and kinematics of the mission environment. The simulation takes a polyhedron model of the asteroid, a third-body disturbance by the sun, and an additional acceleration due to solar radiation pressure into account. The simulation forms a base for the sensor measurement simulation. As the IMU consists of an accelerometer and a gyroscope, the measurements total to two sets of available data for position as well as attitude estimation.
The navigation filter estimates the position, velocity, attitude and gravitational constant of the asteroid, by use of an extended Kalman filter (EKF). The EKF is augmented for the quaternion states to a multiplicative extended Kalman filter. The navigation filter is simulated for a benchmark scenario, as well as for different orbital heights and temporary loss of the star tracker as well as the LiDAR sensor.

As a result, it is concluded that it is feasible with the given sensor set to approximate the position and attitude of a spacecraft in proximity of 433 Eros. For the position, a root-meansquare error (RMSE) of 0.5 m is found at an orbital height of 50 km. Using a time step of 0.1 s for the EKF is recommended after a trade-off between accuracy and computational time. It is concluded that the proposed model for state estimation is sufficiently accurate for position and attitude estimation for the given benchmark scenario. With this navigation filter, we come one step closer to the development of autonomous navigation for asteroid observing spacecraft.