As industry increasingly relies on automation, the need for accurate and reliable inspection, maintenance or transportation methods becomes crucial to ensure operational integrity. Mobile robotics is positioning itself as a discipline where a robot moves and performs such tasks, although sometimes there are complex environments, in which the cognitive information necessary for successful execution is lacking. The Thesis presents the results of research aimed at improving such processes, in an automated way, in real environments of three complex industrial sectors. The thesis investigates the specific reasons why certain tasks, in the proposed scenarios, present a complexity greater than and sometimes intractable than that which can be solved with conventional sensors. Solutions are developed in the three use cases and advances are presented on the current state of the art that, including detectors based on single-stage artificial intelligence models, allow to address and improve the perceptual processes of the robots under study. In this way, the contributions in the Thesis present improvements in these processes through the inclusion of YOLO algorithms as an advanced perception sensor.