Neuronics ADAS Autonomy Box, a compact edge-AI compute module with two round connectors, shown alongside an off-road UGV and a robotic quadruped on a dark tactical background
    Back to all solutions

    ADAS & Autonomy for Ground Robotics

    AUTONOMY BOX

    Off-road autonomy for UGVs and quadrupeds

    A sealed compute module that turns a teleoperated ground platform into an assisted or fully autonomous one. Built for unstructured terrain rather than road lanes: the perception stack expects slopes, vegetation, ruts and dust, and it adapts to both wheeled or tracked UGVs and legged quadrupeds.

    ADAS layer

    • Obstacle detection with negative-obstacle and ditch awareness
    • Collision avoidance and speed governing under operator control
    • Rollover and slope-limit protection for off-camber terrain
    • Assisted teleoperation over degraded or high-latency links

    Autonomy layer

    • Waypoint and route following across unmapped terrain
    • Traversability estimation and cost-map planning
    • Follow-me, leader-follower and convoy behaviors
    • Return-to-operator and comms-loss fallback behaviors

    Navigation

    • GPS-denied localization with visual-inertial odometry and LiDAR SLAM
    • Geo-registration against prior imagery where available
    • Terrain-relative pose estimation for legged gait planning
    • Area and building scanning with mapped output

    Platform fit

    • Off-road UGVs — wheeled and tracked, from light mules to heavy carriers
    • Quadrupeds — gait-aware planning and step placement for robotic dogs
    • Vehicle-agnostic drive-by-wire and actuator abstraction layer
    • Swarm and hybrid-swarm logic coordinating UGVs, quadrupeds and drones

    Interoperability

    • ATAK / TAK Cursor-on-Target reporting for dismounted operators
    • STANAG 4818 interoperable control and STANAG 4609 imagery pass-through
    • ROS2, CAN, serial and I2C buses for sensors, payloads and vehicle ECUs
    • Optional onboard ATR detector-tracker for ISR and C-UAS overwatch tasks

    At a glance

    Compute
    NVIDIA Jetson (Orin family), Qualcomm
    Sensors
    Stereo, LiDAR, IMU, EO/IR
    Platforms
    Wheeled, tracked, quadruped
    Interfaces
    ROS2, CAN, serial, I2C, custom
    Standards
    STANAG 4818, STANAG 4609, ATAK / CoT
    Teaming
    Swarm and hybrid-swarm coordination
    Navigation
    GNSS and GPS-denied
    Origin
    Designed and supported in Israel
    NVIDIA JetsonQualcommiMX8Hailo

    Integration FAQ

    Which ground platforms can the Autonomy Box be fitted to?
    Wheeled and tracked off-road UGVs, from light mules to heavy carriers, as well as robotic quadrupeds. A vehicle-agnostic drive-by-wire and actuator abstraction layer handles the platform-specific control interface.
    Does the Autonomy Box work without GPS?
    Yes. It localizes with visual-inertial odometry and LiDAR SLAM, and can geo-register against prior imagery where available, so navigation continues in GPS-denied and jammed environments.
    What sensors and interfaces are required?
    Typical builds use stereo cameras, LiDAR, an IMU and optional EO/IR. The module connects over ROS2, CAN, serial, I2C or a custom bridge to the existing vehicle control stack.
    Can operators keep manual control?
    Yes. The ADAS layer runs underneath teleoperation, providing obstacle and negative-obstacle detection, collision avoidance, speed governing and rollover protection while the operator drives, including over degraded or high-latency links.
    Comms-loss fallback behaviors take over, including return-to-operator and safe-stop. Waypoint and route following continue across unmapped terrain using onboard traversability estimation and cost-map planning.
    Does the Autonomy Box support swarm or hybrid-swarm operations?
    Yes. Multiple UGVs and quadrupeds can operate as a swarm with shared cost maps and task allocation, and a hybrid-swarm can mix ground platforms with drones so one operator supervises the whole team from a single control surface.
    How does it integrate with ATAK and NATO standards?
    Vehicle position, mapped areas and detected objects are published as Cursor-on-Target messages to ATAK / TAK clients. Control interoperability follows STANAG 4818, and onboard video can be passed through as STANAG 4609 motion imagery with KLV metadata.
    Can it run ATR for C-UAS or ISR overwatch?
    Yes. The same Neuronics ATR detector-tracker can run on the box, giving a UGV or quadruped onboard target recognition for ISR and C-UAS overwatch while the autonomy stack continues to drive the platform.

    Integrate AUTONOMY BOX into your platform

    Tell us your compute, sensors and platform, and we will scope the integration.