Our ELROB 2026 robot, an outdoor logistics build. The Outdoor Navigation Package is the front sensor frame. The cargo box, the Follow Me module and the custom colour are separate options and are not part of it.
Outdoor Navigation Package
Panther and Lynx already handle the terrain. This package lets them drive the route without an operator.
One front-mounted assembly carries dual-antenna RTK GNSS, a 360° 3D LiDAR and a depth camera. Husarion's sensor fusion keeps localisation continuous as the robot moves from open sky into GNSS-denied ground, so there is no mode switch and no hand-off to the operator when it drives between buildings or under tree cover.
You do not integrate any of it. The robot arrives with the package mounted, calibrated and validated, and the mission UI runs in a browser on the robot itself.
1 cm
RTK position accuracy
2
GNSS antennas for true heading
360°
3D LiDAR coverage
10 cm
route repeatability
0
cloud services required
120 kg
payload still free on Panther
The problem it solves
Outdoor autonomy tends to fail on integration rather than on algorithms. Three failures come up again and again.
A single GNSS antenna gives position but no heading. A magnetometer sitting next to four BLDC motors and a steel chassis drifts badly, and gyro or wheel odometry drifts over time, so the robot ends up confidently facing the wrong way.
RTK is excellent under open sky and useless between buildings or under dense canopy. Systems that switch between GNSS and LiDAR modes tend to lurch at the boundary, which is usually where the route is tightest.
Then there is everything else: correction services, satellite map layers, time synchronisation, extrinsic calibration, mounting that survives vibration. Each one is solvable on its own, and together they are the reason pilot projects slip by months.
This package covers all three as one delivered configuration.
What's on the robot
The package is the front sensor frame and the computer behind it. Everything else is the standard platform, or equipment you choose separately.
The Outdoor Navigation Package on a standard Panther, in standard livery. This is what the package on its own looks like.
The same package on our ELROB 2026 robot, with the components labelled. Note that this machine is a full outdoor logistics configuration: the cargo box, the Follow Me module and the custom colour are options ordered separately, not part of the Outdoor Navigation Package.
| Component | Part | Why this one |
|---|---|---|
| GNSS antennas | 2 × NovAtel VEXXIS GNSS-850 | Multi-constellation, multi-frequency. Phase centre stability under 2 mm, patented multi-point feed with multipath rejection, low-elevation tracking for tree lines and urban canyons. MIL-STD-810G shock and vibration, IP69K sealed, −40 °C to +85 °C |
| GNSS receiver | u-blox ZED-F9P | L1 / L2 / L5, RTK accuracy 0.01 m + 1 ppm CEP, convergence under 10 s. Moving-base mode with the two antennas produces heading directly |
| 3D LiDAR | Livox Mid-360 | 360° horizontal coverage for localisation and obstacle detection where GNSS cannot reach |
| Depth camera | Luxonis OAK-D Wide | Close-range perception in front of the robot |
| Connectivity | Teltonika PR1KC640 combo antenna | Mobile internet for correction services and remote access |
| Operator interface | Browser UI, hosted on the robot | Mission planning from any laptop on the network. An on-robot touchscreen is available as an option |
Two antennas on a fixed baseline let the receiver compute absolute heading from the geometry between them. It is available the moment the robot powers on, it does not drift, and magnetic disturbance from motors, chassis or ferromagnetic terrain does not affect it. On a robot that spends much of its time stationary or moving slowly, that is what keeps its reported orientation trustworthy.
Mission modes
| Mode | What the operator does |
|---|---|
| Navigate Through Poses | Continuous path tracking. Draw the path on the map, or record it by driving the robot manually once, then replay it as often as the mission needs |
| Navigate Through Waypoints | Sequential multi-stop missions with configurable dwell time at each point |
| Navigate to Pose | One click to a target pose |
Add the Follow Me tether module and the operator teaches the route on foot, including at night or in dust where camera-based teach-in degrades. The robot then replays it. This is the workflow we ran at ELROB 2026.
Follow Me is currently offered through our Field Evaluation Program rather than general order. Route teach-in by driving the robot manually is available today, and works with any of the mission modes above.
How accurately does it repeat a route?
On a route replayed under good RTK conditions, the robot tracks the taught line to within about 10 cm laterally. That is the figure that matters for planning: it tells you whether a route fits down a service road, past a loading bay, or between two parked vehicles.
Note that this is a different number from the 1 cm RTK position accuracy quoted for the GNSS receiver. The receiver knows where it is to a centimetre. The robot then has to steer a physical vehicle along a line, over terrain that pushes it around, so path error is larger than position error. Anyone quoting sensor accuracy as path accuracy is quoting the wrong number.
Deviation grows where RTK degrades, on loose or heavily cambered surfaces, and at higher speed. We tune for your terrain during commissioning.
Indoors and under canopy, with no prior map
Where GNSS drops out, the robot builds its map as it goes and localises against it live. There is no separate mapping run to schedule, no site survey to pay for, and no map file to maintain as the site changes.
In practice this means you can drive from an open yard into a building, through it, and back out, and the robot stays localised the whole way. The transition is handled by the sensor fusion rather than by an operator switching modes.
Mission UI, hosted on the robot
Point a browser at the robot's IP address and plan on a live map view. No cloud dependency. No external account. No vendor lock-in. Once correction data is available locally, the robot stays fully operable on an isolated network with no internet at all.
Your code and ours run on separate computers
The package runs on its own dedicated computer, closed and doing one job. Your application runs on the standard user computer inside Panther or Lynx, exactly where it would sit on a robot without the package. The two talk over a shared subnet, and the package exposes its ROS 2 API across it.
That separation is deliberate. Your perception stack cannot starve navigation of CPU, a crash in your code does not take down localisation, and you can redeploy your application without touching a system that has been calibrated and validated. You get the ROS 2 interface to command missions and read state. You do not get to modify the navigation stack itself, which is what keeps it something we can support.
Built-in safety
The robot stops automatically when it detects an obstacle in its path. A software E-stop can be triggered and reset from the UI. Panther and Lynx also keep their own hardware E-stop on the robot, independent of any package.
The package provides operational safety functions. It is not a certified functional-safety system and carries no performance level or safety integrity level rating. Deployments around people require your own site risk assessment.
Field record
Again the ELROB logistics build. The navigation package is the sensor frame at the front.
At ELROB 2026, the European Land Robot Trial, we ran this package on Panther in the Transport-Mule scenario: a shuttle mission between two points roughly 300 m apart in non-urban hill terrain in Switzerland, alongside 21 teams including Diehl Defence, Rheinmetall, L3Harris and TNO. The route was taught by walking it with Follow Me, then replayed autonomously.
One full shuttle cycle completed, with one terrain-related safety intervention on a steep uneven section. A second cycle was stopped by a barricaded gate. Dynamic replanning around blockades of that kind was outside the scope of the test, and it is outside the scope of this package today.
We publish the intervention because you are likely to meet the same terrain.
Where it fits, and where it doesn't
It is designed for repeatable outdoor missions on structured and semi-structured terrain: gravel and service roads, industrial sites, campuses, farms, test ranges, logistics paths, and the transitions between outdoor and indoor space.
Ask us first if your mission involves highly unstructured off-road driving, routes that are frequently blocked and need complex dynamic replanning, or environments with neither usable GNSS nor stable LiDAR structure. None of these rules the package out, but they need a use-case review before anyone commits to a deployment date.
What we do, what you provide
We deliver the robot with the package mounted, calibrated and field-tested, the mission UI running on it, the configuration tuned to your terrain during commissioning, and the full manual and setup documentation.
You provide the site and these three things:
| Mobile internet or local network | To reach an NTRIP correction service |
| NTRIP correction service | Strongly recommended in open terrain, because centimetre-level RTK depends on it. Without corrections the system still runs, at reduced absolute accuracy |
| Satellite map layer | Mission planning context in the UI |
Common questions
Can I add this to a Panther or Lynx I already own? Yes. The robot comes back to us for mounting and calibration, then ships back to you. Factor in shipping both ways and the downtime.
How closely does it follow a taught route? About 10 cm laterally under good RTK. See the section above for what widens that.
Does it need internet to run? It needs a correction source for centimetre-level RTK, which normally means mobile internet reaching an NTRIP service. Once corrections are available locally, the robot runs on an isolated network with no internet at all. Nothing about mission planning or execution depends on a cloud service.
Does it work indoors, and do I have to map the site first? Yes, and no. It builds its map live and localises against it, so there is no mapping run to schedule and no map file to keep up to date as the site changes.
Can I run my own ROS 2 code on the robot? Yes, on the robot's standard user computer, reaching the package over the shared subnet through its ROS 2 API. Your code and the navigation stack run on separate computers, so neither can destabilise the other.
Is the robot slower or more limited when driving itself? No. The platform's own speed and slope limits apply, with no autonomy-specific derating.
Is this a certified safety system? No. It stops automatically on detected obstacles and offers a software E-stop, and the platform keeps its independent hardware E-stop, but there is no performance level or safety integrity level rating. Deployments around people need your own risk assessment.
What if my terrain is rougher than what you describe? Talk to us before committing to a date. See where it fits, and where it doesn't. A use-case review costs nothing and is faster than a failed pilot.
Compatibility and ordering
| Platforms | Panther, Lynx |
| Order code | NAV01P on Panther, NAV01L on Lynx |
| Mounting | Front assembly. The user compartment and the space above it stay free for your payload, up to 120 kg on Panther |
| Combines with | Follow Me and the standard configuration options |
| Delivered | Mounted, calibrated and field-tested by Husarion |
| Lead time | Typically 4 to 8 weeks, depending on configuration and current queue |
| Retrofit | Yes, to a Panther or Lynx you already own. The robot returns to us for mounting and calibration, so allow for shipping and downtime |
| Autonomous limits | Platform limits apply. No autonomy-specific derating of speed or slope |
| Warranty | 1, 2 or 3 years, per the standard warranty options |