The global positioning system (GPS) is vulnerable to jamming and becomes inoperable in harsh environments such as urban canyons, metal buildings, tunnels and mines. In 2018, the Ministry of National Defense challenged the country’s scientific community to propose solutions to achieve the positioning of a soldier in a resilient manner, without resorting to satellites.
The overall objective of the project was to provide a proof of concept for a navigation system that is:
• Python: Cython, Numpy, Matplotlib, Pandas
• DeepMimic
• C++
• Keras
Skadra has proposed a system based on inertial navigation, which analyzes human movement to measure the distance of each step and changes in orientation.
The inertial measurement units are mounted directly on the soldiers’ boots, and the data is recorded on a Raspberry Pi single board computer. The GPS signal is also acquired in parallel to initialize the system at a known position and provide a reference against which the Skadra system can be compared.
Over 20 hours of walking (approx. 100 km) were recorded, providing an analysis of the reproducibility and stability of the results.
Finally, a biomechanical gait model was developed to study different approaches to simulation and to train a processing algorithm based on machine learning.
LTI’s prototype enables positioning to an accuracy of a few tens of meters after 30 minutes’ walking on routes including tunnels, stairs and outdoor sections.
Development is continuing in-house to increase the repeatability of results and to determine the benefit of using artificial intelligence-based processing compared with more conventional methods.
