AI applied to safety in indoor climbing

Autobelay devices are widely used in climbing gyms and allow solo climbing: the rope retracts automatically during the ascent and, in the event of a fall, the system brakes and lowers the climber in a controlled manner. In addition, they reduce human errors in belaying and facilitate training.

In this context, the inLab FIB at the Universitat Politècnica de Catalunya – BarcelonaTech (UPC) has developed for the company Lizcore® a machine learning–based proof of concept applied to its autobelay systems. The objective is to improve safety before starting the climb by automatically, reliably and accurately detecting potential risk situations that could cause accidents.

The AI model is integrated into the software of Lizcore’s Safety Gate, a system that automatically monitors the use of autobelay devices through artificial intelligence and computer vision techniques. This system enables real-time alerts and, thanks to its direct connection with the management software (Lizcore GymOS), facility managers and technicians can receive immediate notifications and access precise usage data to manage preventive maintenance.

This innovative device automatically checks:

• Access control: Only authorised climbers with up-to-date legal agreements can activate the autobelay device using the NFC wristband.
• Intelligent verification (AI): A front-facing camera validates the correct connection of the harness and the locking of the carabiner to the ventral attachment point before allowing the activity to begin.
• Route safety and range of action: A top-mounted camera verifies that the climber remains within the correct route and within the safety margins provided by the autobelay line.
• Landing zone control: A front-facing camera monitors the fall or descent area to prevent collisions with other people who may be crossing or positioned beneath the climber.
• Multi-channel alert system: In the event of any malfunction or dangerous situation, acoustic warnings, visual signals (LEDs) and immediate mobile notifications are activated for the facility staff.

This proactive safety layer complements —rather than replaces— human supervision, providing a key technological tool to prevent accidents caused by distraction or human error.

The system implemented by the inLab FIB uses machine learning models based on convolutional neural networks (CNNs), trained with images from the Safety Gate cameras. These models enable:

• Identification and localisation of objects (harnesses, carabiners, devices).
• Recognition of actions and behaviours (clipping in, starting the route, attempting to climb without being secured).
• Detection of risk patterns in real time.

The solution developed by Lizcore® with the support of the inLab FIB enables:

• Reduction of accidents caused by human error through AI.
• Provision of real-time alerts and prevention.
• Improvement of the climbing experience with greater safety and confidence.
• Provision of useful data for the management and maintenance of climbing gyms.
• Promotion of digitalisation and technological innovation in the indoor climbing sector.

Lizcore® is a Catalan technology startup that applies artificial intelligence to the world of indoor climbing to improve safety, performance analysis, sport gamification and climbing gym management.