RB-WATCHER: Artificial Intelligence Applied to Vision-Based Recognition

The RB-WATCHER, Robotnik's Autonomous Mobile Robot for inspection, integrates Artificial Intelligence through vision-recognition modules, enabling the robot to perceive and analyze its surroundings autonomously. These systems form the core of its detection and image-processing capabilities and represent one of the robot's main development lines in 2025.

Vision recognition architecture
The RB-WATCHER's vision system is designed to continuously process the images captured by its cameras. When it recognizes an element that matches previously trained patterns, a detection is generated.

For example, in the case of person recognition:
• Image capture: the robot's cameras take real-time images of the environment.
• Processing and classification: the AI compares image features against previously trained models to determine whether they correspond to a person or another object.
• Subsequent action: once detected, the information is recorded in the database, the detected video or image is saved, and an alert is sent.

Training these modules requires a large and representative dataset. For instance, a training simulation for person detection may involve tens of thousands of images that teach the system to distinguish between people and other elements in the environment, thereby improving model accuracy.

Advanced industrial inspection capabilities
RB-WATCHER features several vision-recognition modules aimed at inspection applications:
• Person detection: general identification of human presence, with the option to include real-time privacy filters.
• Thermal anomaly detection: using thermographic cameras, the robot can identify unusual heat sources, comparing signals with the environment to assess potential risk.
• Infrastructure monitoring: detection of damaged fences, vehicles, or license plates, logging events in databases and generating automatic alerts.

Extensibility and training of new modules
RB-WATCHER allows the development and training of new recognition modules tailored to specific needs. This includes:
• Training for the identification of industrial indicators, such as variations in nanometers, pipe drips, visible leaks, loss of temperature in components requiring constant thermal conditions, as well as other elements specific to industrial environments — including the use of personal protective equipment (such as helmets or safety boots) and detection of critical components.
• Adaptation for specific clients or projects, such as systems for gas detection or safety-condition monitoring in industrial plants.

The development process of these modules follows a cycle of data and image collection, labeling, model training, validation, and continuous tuning to ensure accuracy and robustness in real environments.

AI integration with robot operation
One of the most significant software advances implemented in the RB-WATCHER is the integration of vision modules with the Robot Management System (RMS), an autonomous system responsible for managing the execution, prioritization, and cancellation of missions based on the robot's various operational parameters.

The RMS operates independently from the alarm system and makes decisions based on factors such as battery level, sensor and camera status, the duration and number of pending missions, and time spent in an idle (IDLE) state.

Among its main functions:
• Event notifications related to mission execution: start, completion, cancellation, and encountered errors.
• Time management: controls the maximum time the robot can remain IDLE (an indeterminate situation that may be due to obstacles, navigation issues, or blockages). If this limit is exceeded, the RMS commands an automatic return to the charging station (docking).
• Battery management: defines different energy thresholds that determine the robot's operational behavior:
• Sufficient level to start missions.
• Minimum level to accept and execute missions.
• Critical level to reject missions and return to docking. Once docked, it manages the charge increase required to resume operation, depending on the number of pending missions, time, or battery percentage reached.
• Advanced energy-safety management: the RMS includes additional protection measures:
• Automatic shutdown when battery reaches critical level.
• Notification and hibernation if the robot cannot return to its charging station properly.
• After starting with a critically low battery, the system grants a prudent amount of time to reach the docking station, varying according to the operating mode (automatic or manual).

The integration of AI, visual perception, and autonomous mission management makes the RB-WATCHER a highly intelligent system with enhanced capabilities that translate into user benefits.
For example, detecting overheating points in electrical panels, unauthorized human presence in restricted areas, leaks or drips in pipes, damage to fencing, poorly positioned vehicles, or lack of personal protective equipment. These cases demonstrate how AI integration improves accuracy, autonomy, and responsiveness in critical industrial environments.