Pendulum Navigate for the Dismounted User

Electronic warfare now dominates 

‍

GPS has become target alpha across all current and potential fields of conflict. Electronic warfare – in the form of jamming and spoofing – means that traditional GNSS relays are no longer an assured form of geolocation. 

This is having a profound effect on how the warfighter operates in contested environments. With the development of more secure M-Code satellites still years behind schedule, armed forces are turning to alternative positioning, navigation, and timing (alt-PNT) solutions to protect assets and personnel. These can be broadly bucketed into hardware- and software-based solutions. Existing hardware fixes include quantum computing, inertial navigation systems, and inertial measurement units. The solutions themselves are medium- to high-value assets, so largely remain the reserve of ‘exquisite’ systems. 

Their high SWaP specifications make them suitable for aircraft, land vehicles, larger UAVs, and other established PoR assets – but unfeasible for individual warfighters and the consumer-grade (COTS) devices that they use in the field. They also remain largely unscalable – at a relatively high price point, one asset per end-user device is not feasible across larger units and divisions. 

Where a software-led approach has been adopted for alt-PNT – solution developers are primarily building capabilities for data-rich drones, as well as land vehicles and seafaring vessels, where computer vision from on-board cameras is readily available. 

All this has led to a crucial divide in the field: how do we protect the 14 million dismounted users across the United States and allied nations who depend on reliable PNT? 

Introducing Pendulum Navigate 

‍

Pendulum Navigate is a software-defined alternative PNT solution. It's built for the consumer-grade devices warfighters’ use as part of their everyday loadout. 

Unlike existing alt-navigation solutions, it does not require any hardware, wiring, or dongles. All data processing is handled on the end-user device, completely removing the need for cloud, GPS, or any other external connectivity. It currently integrates with smartphones, smartwatches, and lightweight drones (sUAS).

Pendulum Navigate combines inertial sensor inputs with preloaded geospatial data and families of lightweight machine learning models to produce a continuous, accurate geolocation output on the end-user device. 

A hardware-agnostic solution, it is designed to integrate with existing navigation and mapping applications and to be available throughout the warfighters’ COTS device ecosystem. 

‍

‍

‍

Unlike other software-based approaches to PNT – Navigate can operate on end-user devices in jammed and spoofed environments without computer vision and LiDAR-based inputs. Here’s how: 

‍

Inertial Sensors 

‍

Pendulum Navigate’s primary inputs are inertial sensor data. These inputs, specifically the magnetometer, barometer, gyroscope, and accelerometer, are found on all consumer-grade devices. Drones also have additional sensors such as altimeters. 

‍

‍

Alone, these inertial sensors are far too noisy to ensure accurate geolocation; but Pendulum Navigate creates a positive, continuous output by combining them with additional data sources and machine-learning models. 

‍

Geospatial Data

‍

Geospatial data can be found for almost everywhere on the planet. Pendulum leverages preloaded digital terrain elevation data (DTED) in the modeling process. While the resolution of this data varies on the location – there is both consistent and continuous mapping across most geographies. 

‍

‍

‍

DTED and land cover data also differs from digital surface models, making it more suitable for generating dismounted outputs. Additional land cover data also improves the overall model accuracy. 

‍

Machine Learning Modelling  

‍

Pendulum processes these two inputs through a family of machine learning models. Running this kind of AI at the tactical edge has long been considered too computationally heavy for COTS devices; however, Pendulum’s approach uses lighter-weight models to significantly reduce the compute burden on the end-user device. 

That means they can run over a significantly longer time and distance in more remote, offline environments with minimal extra power draw. 

‍

Geolocation Output 

‍

Navigate primary output is a continuous point of geolocation (eg. a latitude and longitude prediction, or equivalent measurement). The solution embeds within the Android Team Assault Kit (ATAK) – the trusted geospatial application used by United States armed forces and first responders – users can toggle to Navigate in-app in just a few clicks. 

‍

‍

‍

The video above shows Navigate in use around a technical course. GPS is disabled from the EUD (smartphone), the solution uses only the preloaded data, inertial sensors, and ML to continuously geolocate. The average error rate in this exercise is less than 20 meters. 

The ATAK user also has the option of displaying a heatmap or elliptical circumference. This is a visual representation of the confidence interval of the output. It remains dynamic for as long as they are using the application – which means it can grow and shrink depending on the data available. 

Always-On, Always-Accurate 

‍

Pendulum Navigate processes all this data on device – completely excluding the need for any external connectivity. That not only means it remains operable in jammed and spoofed environments, but also in environments vulnerable to indirect interference (canyons, mountainous regions, heavy canopy cover). 

Whereas existing inertial navigation systems (INS) use similar data points, they process them via a Kalman filter. Kalman filters tend to work better in linear travel environments – and produce a compounding error over distance. By contrast, Pendulum’s ML modelling uses a particle filter. This produces no compounding error for the end user – instead they have a much more dynamic confidence interval which changes depending on the available data at any given location point. 

As noted above, this approach can also operate independently of computer vision or LiDAR data on land. Up until now, this has been used as a crucial data point for reference matching with the available on-device data. Removing the need for camera data not only widens the range of end-user devices the solution is compatible with, but also vastly reduces the level of power draw. For the dismounted warfighter, that means being able to use the application for longer, in more extreme environments, and much more interoperably with their existing navigation applications. 

Protect the Warfighter

‍

The war in Ukraine first highlighted how pivotal consumer-grade devices like smartphones and sUAS will be to the future of active warfare. That same future will also be dominated by an unprecedented level of GPS interference.  At Pendulum, we’re equipping these same devices with always-on, always-accurate geolocation – giving the dismounted user a crucial source of PNT, all within just a few clicks. 

‍

To find out more about Navigate, or to see a demo, reach out to the Pendulum team at navigate@pendulum.global 

‍