LiDAR safety
360° perception to operate among people, furniture and real routes.
A technical and visual explanation of the navigation and safety system: LiDAR, stereo vision, ultrasonics, reaction zones, braking and local logic even offline.
KEENON LiDAR system: autonomous navigation and safety in real environments.
Visual guide to the multimodal navigation system: LiDAR, 3D vision, safety zones, obstacle reaction and protection in real environments.
Multimodal sensor fusion
KEENON robots combine multiple perception technologies simultaneously. No sensor works alone: the central AI system fuses all data streams in milliseconds to build a real-time 3D map of the environment and make safe navigation decisions.
How does the LiDAR system work?
LiDAR (Light Detection And Ranging) emits high-frequency laser light pulses and measures the time each pulse takes to bounce back off an object. With this information, the robot builds a precise 3D point cloud of the environment, detecting obstacles at distances from centimetres to several metres.
240° LiDAR (T-series robots and C30)
240-degree front and side coverage. Detects static and dynamic obstacles in the horizontal plane at floor level.
360° Dual LiDAR (S100 / S300 / C40 / C55)
Two combined LiDAR units provide full 360° perimeter coverage, with no horizontal blind spots.
VSLAM + LiDAR Fusion
LiDAR is combined with Visual SLAM (Simultaneous Localization And Mapping): the cameras reinforce precise localisation without the need for physical markers.
The 10 detection components (C40 / C55)
The C40 / C55 model features the most complete perception set in the range, with ten components working in parallel and coordinated by the central AI engine.
Main LiDAR (horizontal)
360° horizontal scan. Detects walls, furniture, fixed equipment and people at floor level. Range up to 25 m. Update at 10–25 Hz.
Secondary LiDAR (low/high)
Covers angles not reached by the main LiDAR: low areas (feet, obstacles at floor level) and raised areas suspended above the floor.
3D stereo camera — front
High-resolution stereo pair facing forward. Generates a depth map to detect complex shapes, moving people and suspended objects.
3D stereo camera — side
Monitoring of the left and right flanks. Essential for narrow aisles where the robot needs millimetre navigation margins.
Line laser — high precision ×4
Four line lasers project horizontal detection planes at different heights. They capture very low obstacles (< 5 cm) that standard LiDAR might miss, such as cables, steps or liquids.
Ultrasonic modules ×3
Ultrasonic transducers that emit sound waves and measure the echo. Effective with transparent surfaces (glass, plastic) and in low-light conditions where optical sensors may fail.
Emergency Sensing Strip (contact strip)
A sensor strip that surrounds the base of the chassis. If direct physical contact occurs with any part of the robot (accidental collision, deliberate act), it immediately triggers the emergency stop.
Physical emergency stop button
A physical button accessible to staff. It stops all motors instantly and irreversibly until manually reset. Complies with industrial safety regulations.
IMU (Inertial Measurement Unit)
Gyroscope and accelerometer. Detects tilts, sharp knocks, tipping attempts or forced pushing. Complements the information from the optical sensors.
Wheel odometry encoders
They measure the exact rotation of each wheel. They allow the system to detect whether the robot is being forcibly stopped (wheel slip or blocking) and react accordingly.
Concentric protection layers
The robot defines dynamic safety zones around itself. Each zone triggers a different system response depending on the proximity of the detected obstacle.
🔴 Zone 1 · < 0.4 m
Instant emergency stop. The motors stop in milliseconds. The robot does not resume until the zone is clear.
🟠 Zone 2 · 0.4 – 1 m
Maximum deceleration and activation of an audible alert. The robot prepares to stop if the obstacle is not removed.
🟢 Zone 3 · 1 – 2 m
Preventive speed reduction. The AI engine starts calculating an alternative route in parallel.
🔵 Zone 4 · 2 – 4 m
Active monitoring. Tracking the obstacle to predict its trajectory and plan detours in advance.
From sensor to action in milliseconds
When a sensor detects a threat, the processing chain operates in a pipelined way: each stage runs in parallel without blocking the next, guaranteeing response times below 5 ms.
Detection by primary sensor
LiDAR / stereo camera / ultrasound detects a new object in its field of view. The raw data (distance, coordinates, angle) is sent to the internal data bus.
AI fusion and classification
The central AI engine combines the data from all active sensors, classifies the obstacle (person, fixed object, moving object) and updates the 3D map of the environment in real time.
Zone and speed evaluation
The system determines which safety zone the obstacle is in and calculates the maximum safe speed. Simultaneously it explores alternative routes if any are available.
Action on the motors
The command is sent to the motor controller: reduce speed, stop or divert depending on the zone. In the case of zone 1 it also activates the horn/LED visual alert.
Triple emergency stop (if applicable)
If physical contact occurs despite everything (chassis sensor strip activated or button pressed), the three stop systems act simultaneously: software control, redundant electrical cut-off and mechanical brakes.
Notification and resumption
The event is logged with a timestamp, obstacle type and location on the map. The IoT dashboard (app / web) receives the alert. When the zone is clear, the robot resumes the mission automatically.
Defence mechanisms against malicious actions
KEENON robots are not only designed to avoid accidental collisions: the system incorporates multiple layers of physical, electronic and operational protection to respond to attempts at sabotage, vandalism or theft of the equipment or its load.
Design principle: The robot interprets any unexpected physical action (pushing, blocking, tipping attempt, unauthorised access to compartments) as a security event and triggers the corresponding response chain.
Unauthorised contact detection
The Emergency Sensing Strip detects any unexpected physical contact with the chassis — including knocks, shoves or attempts to hold the robot back — and triggers an immediate stop with an alert to the central system.
IMU monitoring — impact detection
The inertial unit records vibrations, sharp knocks and tipping attempts. Any anomalous acceleration generates an alarm event and is logged with GPS coordinates from the indoor map.
Password-protected compartments
The T3 and W3 models feature password-protected locks on their cargo compartments. Unauthorised access is logged and can trigger remote alerts.
Real-time alerts (IoT Dashboard)
All security events (stops, impacts, access attempts, unexpected disconnections) are sent to the IoT dashboard and the operator's mobile app with a timestamp and map position.
Vision cameras with recording
The stereo cameras operate continuously during the mission. The images captured at the moment of a security event can be used as evidence in vandalism incidents.
Audible and visual alarm
In a security event, the robot activates its LED indicators and horn to deter the aggressor, alert nearby staff and make the emergency situation visible in the surroundings.
Secure offline operation: The robot operates completely autonomously without relying on Wi-Fi or 5G. This means an attack on the local network cannot disable its safety systems or take remote control of the robot.
Perception set by range
| Model | LiDAR | Stereo vision | Line lasers | Ultrasonics | Emerg. stops |
|---|---|---|---|---|---|
| C40 / C55 (cleaning) | Dual 360° | 2 × stereo | 4 units | 3 modules | 3× redundant |
| S300 / S100 (heavy load) | Dual LiDAR | Tri-stereo | — | Included | 3× redundant |
| T10 / T11 (restaurant) | VSLAM + LiDAR | 5 × stereo | — | Included | 2× + strip |
| T3 (enclosed delivery) | 240° LiDAR | 3 × stereo | — | Included | 2× + strip |
| C30 (medium cleaning) | 240° LiDAR | 3 × stereo | — | Included | 2× standard |
| W3 (hotel) | Laser+Visual SLAM | 120° wide-angle | — | Included | 2× standard |
Why the system is safe by design
Multiple redundancy
No sensor works alone. If one fails, the rest maintain safety coverage. The emergency stops are triple and independent of each other.
Millisecond response
The AI engine operates in a continuous closed control loop. The latency between detection and action on the motors is under 5 ms in normal conditions.
Real-time dynamic map
The environment is not static. The robot updates its map on every cycle, remembering where obstacles were and predicting the trajectories of moving objects.
No network dependency
It operates completely offline. A Wi-Fi outage, a network attack or interference does not compromise safety: all the logic runs locally on the robot.
Note for integrators: The safety-zone thresholds, the reaction speed and the sensitivity parameters can be configured from the KEENON management software according to the specific needs of each facility — hospital environment, supermarket, hotel, logistics warehouse, etc.