Bringing dynamic route intelligence to UGVs at the edge

Illustration generated with AI

Uncrewed ground vehicles are growing in importance in military operations and disaster response. They can reduce risk to personnel, extend operational reach and perform tasks in areas that may be difficult or dangerous for humans to access.

However, as operations in Ukraine have clearly demonstrated, operating conditions can change quickly and it is often impossible to rely on persistent communications or pre-planned routes.

Obstacle avoidance is not enough

For route planning, this presents a challenge. Modern autonomy stacks such as ROS 2 Nav2 are highly capable of obstacle avoidance, local planning and route execution. When an obstacle appears, the vehicle can detect it and navigate around it.

However, many systems still treat each obstacle encounter as an isolated event, and retain no memory of it. As a result, the vehicle may avoid the obstacle and turn around to find an alternative route, only to then forget about the obstacle and approach the same area again. The risk is not only slower progress, but also that the UGV can become stuck, repeatedly making the same mistake. The same challenge applies to multiple vehicles operating in the same area, with no way of sharing what they have learned.

Delays, loss of mobility, unnecessary exposure, or repeated attempts to navigate unsuitable terrain can all affect mission success and reduce vehicle survivability. A vehicle that can adapt its route as conditions change is more likely to remain mobile and mission-effective.

What is missing to achieve this is the ability to dynamically recalculate the route at the edge.

How adaptive navigation should work

Importantly, route recalculations must also consider more than physical obstacles to be effective. Terrain conditions, mission objectives and the tactical situation all influence what constitutes the best route.

Ideally, the vehicle should:

1. Detect an obstacle.
2. Determine that it affects mobility.
3. Update the vehicle’s understanding of the terrain, taking the tactical situation and mission objectives into consideration.
4. Recalculate the route.
5. Remember the information for future decisions.

This creates a memory of traversability, allowing the vehicle to learn from the environment instead of repeatedly treating it as unknown.

Bringing advanced terrain routing onboard

To address this challenge, Carmenta’s Terrain Router has been integrated with the ROS 2 Nav2 framework. Through the Nav2 Global Planner interface, Carmenta’s mobility analysis becomes an operational component of the autonomy stack, enabling advanced terrain-aware route planning and real-time geospatial mobility analysis directly onboard UGVs.

When the UGV requests a global plan between the vehicle’s current position and its destination, Carmenta Terrain Router considers the terrain, vehicle profile, tactical factors and mission constraints. This allows routes to be optimized for mobility, mission success, stealth and vehicle survivability.

The generated route is then passed to the Nav2 framework, where local planners and obstacle avoidance systems execute the mission.

Until now, this type of advanced geospatial analysis has been primarily associated with mission planning systems and operator decision support systems. Running these capabilities onboard fundamentally changes their role.

Introducing a memory of traversability

Geospatial data provides an initial understanding of the operating environment, but no dataset can fully represent conditions on the ground.

As the vehicle moves, onboard LiDAR sensors continuously observe the surrounding terrain. Carmenta technology uses these observations to update a historical traversability map, allowing terrain knowledge gathered during operation to influence future route calculations. The updated terrain understanding can also be shared, allowing other vehicles operating in the same area to benefit from that knowledge.

Instead of every platform learning the same lesson independently, knowledge can accumulate across the mission.

Enhancing stealth and vehicle survivability

Terrain conditions are only one part of route planning. The tactical situation and mission objectives also determine what constitutes the best route. Depending on the task, a UGV may prioritize speed, fuel efficiency, safety, concealment or mission urgency.

Carmenta Terrain Router can incorporate tactical information such as no-go zones, enemy viewsheds and movement corridors into route calculations. By guiding the UGV around areas where it risks detection or engagement, this increases both stealth and survivability. When the tactical situation changes, the vehicle can immediately recalculate its route at the edge, without relying on an external routing service.

From static plans to adaptive autonomy

By bringing advanced terrain routing onboard and combining it with continuously updated mobility information, Carmenta technology enables a more adaptive approach to autonomous navigation.

Vehicles can:

• Replan routes without relying on an external routing service.
• Incorporate terrain observations collected during operation.
• Avoid areas previously identified as difficult, dangerous or impassable.
• Adapt global routes as the tactical information or mission changes.
• Build and share a progressively richer understanding of the operating area.

Integrating Carmenta’s Terrain Router with ROS 2 Nav2 as an active component of the autonomy stack delivers more than terrain-aware routing. It’s a step towards experience-based navigation, where routes are determined and adapted by what vehicles learn while operating in the field. In contested and rapidly changing environments, that ability to learn, adapt and share knowledge is becoming an increasingly important contributor to both mission effectiveness and survivability.