SLAM Navigation &
LiDAR Technology

SLAM is the core navigation algorithm that enables inspection robots to build a map of an unknown environment while simultaneously tracking their own position within it. Unlike GPS-dependent systems, SLAM works entirely from onboard sensor data — making it the essential technology for indoor, underground, and GPS-denied industrial environments.

Our robots implement LiDAR-based 3D SLAM with loop closure detection, enabling centimeter-level positioning accuracy in complex industrial layouts. The robot builds a persistent 3D map on first deployment, then navigates autonomously on all subsequent missions using that map as reference.

Light Detection and Ranging (LiDAR) sensors emit laser pulses and measure the time-of-flight of returning signals to generate dense 3D point clouds of the surrounding environment. Industrial-grade LiDAR units on our robots capture millions of data points per second at ranges up to 100m.

Point cloud data is processed in real-time by the onboard edge compute system to extract geometric features, detect obstacles, and update the navigation map. Multi-return LiDAR handles complex environments including wire meshes, grating floors, and transparent surfaces that challenge camera-only systems.

Industrial inspection environments — power substations, underground tunnels, data centers, oil refineries, and enclosed facilities — are typically GPS-denied. Our robots operate with full autonomy in these environments using LiDAR SLAM, IMU fusion, and wheel odometry.

The navigation stack fuses data from LiDAR, inertial measurement unit (IMU), and proprioceptive sensors to maintain accurate position estimates even in featureless corridors, symmetrical spaces, and areas with electromagnetic interference that would degrade GPS signals.

Inspection missions are defined as sequences of waypoints — specific locations the robot must visit, along with the inspection tasks to perform at each location (thermal scan, gas reading, visual check, meter reading). The robot executes these missions autonomously without human intervention.

Waypoints are configured via the cloud inspection platform with a drag-and-drop map interface. Mission schedules can be set to run at specific times, triggered by events, or executed continuously. The robot automatically returns to its charging dock between missions and resumes patrol on schedule.

Real-world industrial environments contain moving obstacles — personnel, vehicles, equipment being moved. Our robots implement multi-layer obstacle avoidance that distinguishes between static map features and dynamic obstacles, navigating around them without mission interruption.

The local planner uses a combination of LiDAR, depth cameras, and ultrasonic sensors to detect obstacles in real-time. When a dynamic obstacle is detected, the robot pauses, recalculates a collision-free path, and continues the mission. If a path cannot be found, the robot reports the blockage and waits for clearance.

All SLAM computation, obstacle avoidance, and mission execution runs on the robot’s onboard edge compute system — no cloud connectivity required for navigation. This ensures reliable autonomous operation even in environments with intermittent or no network connectivity.

The edge compute platform runs a real-time operating system with dedicated processing cores for navigation, sensor fusion, and AI inference. Inspection data (thermal images, gas readings, visual captures) is buffered locally and synchronized to the cloud platform when connectivity is available.