Hybrid Autonomous Global Localization in the Absence of PNT Infrastructure

Typical global positioning methods require a network of multiple satellites for localization. When such a GPS-like infrastructure is unavailable or when its direct line-of-sight communications with Earth are not warranted, a user on the far side of the moon or near the polar regions would need to rely on other autonomous navigation techniques such as celestial positions, local radiometric infrastructure network, altimetry, and terrain-relative measurements, etc. to determine its in-situ position and orientation. This paper proposes a hybrid scheme combining IMU sensors, star trackers (ST), and surrounding terrain outline (STO) for real-time autonomous localization. The lack of oceans and plateaus and the excess of craters and ridges on the moon help make an STO a signature unique to its location. Due to the absence of vegetation, seasons, weather, and artificial human-made structures on the moon, STOs remain unchanged over time. In addition, since all rovers are equipped with onboard panoramic cameras, STOs can be obtained readily and reasonably inexpensive. Some cameras have infrared capabilities, enabling the rover to self-localize in complete darkness or low-lighting conditions. Due to the contrast between the cold sky background, the moon's diffusing hot body temperature, and the unavailability of the atmosphere, an STO can be obtained very crisply and distinguishably. Once collected, the STO can be matched with a database of constructed STOs for a specific region to determine the position and orientation of the user. The STO database is built in advance using NASA's high-resolution digital elevation map (DEM), acquired by the Lunar Orbiter Laser Altimeter (LOLA) instrument on the Lunar Reconnaissance Orbiter (LRO). Our previous investigations show that, with appropriate instrument calibration and performance, the average self-localization errors can be a few meters to tens of meters, and the average orientation errors are within half a degree. For high accuracy results, the terrain outline matching method requires high-resolution DEM, and thus intensive computational effort and large memory storage. On the other hand, recent advanced star trackers can provide the user's attitude and position within a few seconds, even in lost-in-space mode. However, their three-sigma position errors could be as high as tens of meters. Our proposed scheme first employs the ST technique to attain a quick approximate position and then applies the STO method on a high-resolution DEM but small area enclosing the approximate position. This way, we can achieve accuracy and speed by combining the ST and the STO techniques. Our results show that even at the worst tilting and terrain uncertainties and half of the field of view being blocked, the 3σ positioning error for the STO is less than 5m. When 75% of the field of view is obstructed, the 3σ positioning errors never exceed 25m. Most importantly, the STO method works in areas with relatively flat local terrain.