Implementation of an Autonomous ROS-Based Mobile Robot with AI Depth Estimation

Industry 4.0 is one of the biggest industrial revolutions with the creation of new technology and artificial intelligence solutions. In this context, Autonomous Mobile Robot (AMR) is one of the reasons to expand productivity and logistical flexibility due to its adaptability to new environments. However, only recently the AMR has had greater acceptance, much due to its complexity and cost In this sense, this article presents a ROS-based mobile platform to evaluate the possibility of navigation only with monocular cameras. The hardware and software stack structure will be detailed evidencing the most critical components for the success of the implementation of omnidirectional AMR To ensure a reduced cost while maintaining the advanced ability to detect and change the trajectory whenever an object is detected during its path, it is proposed to use an artificial intelligence system to create an estimated depth image from the 2D RGB images. Finally, to ensure its ability to generalize and navigate, the system accuracy and behavior were evaluated and validated in several environments. The main contribution of this work is the assessment and validation of an AMR navigation solution using only Ai-based depth estimation from 2D RGB cameras without performance impact in an embedded computational system.