Monocular SLAM
September 1, 2008 By: Michael G. Farley, Mark D. Chapman GPS WorldAlternative Navigation for GPS Denied Areas
Extensive research has been performed across the navigation industry in recent years to achieve a practical means of providing GPS levels of navigation accuracy under GPS denied conditions. Benefits from a practical, low cost, implementation of this type of technology are dramatic. Small unit military operations in urban environments would experience a significant improvement in speed and safety resulting from greater situational awareness. Civilian first-responder personnel could map and locate injured victims and dangerous areas. Both tactical military vehicles and civilian traffic management systems could tolerate outages from satellite navigation systems.
![]() |
The vast majority of these efforts have focused on the use of inexpensive MEMS-based dead reckoning units to provide position, navigation, and timing (PNT) information. Unfortunately, these approaches suffer from errors in estimating angular rotation and accrue PNT errors in a linear fashion. Small, low-cost implementations are limited to providing short duration benefit before the error accumulations render them useless. Magnetically hostile environments such as light industrial buildings further degrade PNT effectiveness. Attempts to provide useful MEMS-based error correction sources in an acceptable size, weight, and power consumption tend to be limited to techniques that are equally susceptible to distortion. Stride-based correction mechanisms require precise calibration to the individual and are immature when dealing with combat/first-responder body poses.
Rockwell Collins has addressed these issues by challenging the fundamental assumption that MEMS aiding will provide an acceptable solution. Instead, robotic vision algorithm-based navigation is aided by the MEMS. Current robotic vision algorithms such as Simultaneous Location and Mapping (SLAM) are throughput intensive and are not presently practical for the dismounted user. The use of the MEMS provides excellent instantaneous sensor pointing information that reduces the SLAM processing requirements significantly. The use of Feature Constellation Tracking (FCT) algorithm improvements further reduce the processing requirements by allowing intelligent thinning of the features tracked. This approach has allowed the Ghostwalker IR&D project to demonstrate accurate GPS-denied indoor and outdoor navigation without prior knowledge of features or telltale emissions. Project goals are to develop an inexpensive and practical GPS-denied navigation unit for dismounted soldiers, as shown in the OPENING FIGURE of the proposed Ghostwalker GPS-denied navigation and situational awareness system.
Development History
In 2005 Rockwell Collins began exploring non-GPS-based navigation by developing the navigation and guidance equipment for the DARPA Grand Challenge vehicle, Terramax, in conjunction with Oshkosh Truck, Inc.
After completing the Grand Challenge desert course (one of only five to do so), a survey of current GPS-denied technologies was conducted, focusing on the needs of the dismounted soldier. Desiring a nonemitting solution, this work focused on inertial approaches to the problem. Two of the best dead-reckoning modules (DRM) were evaluated against indoor and outdoor surveyed truth tracks. A variety of DRM error aiding sources were evaluated as well. The basic DRM performance has improved from earlier efforts but still is not accurate enough for warfighter use. Indoor performance for the Honeywell DRM5, as shown in FIGURE 1, diverges quickly in a light industrial environment.
![]() FIGURE 1 Indoor DRM performance |
Outdoors, even with the units calibrated to a particular individual's stride, the Honeywell DRM5 and Vectronix CNM modules had significant error over distance traveled, as shown in FIGURE 2. These errors only increased when other motions such as running, turning, or sidestepping were introduced.
![]() Figure 2 Outdoor navigation performance |
1 2 3 4 5 6 7








