Mobile and service robots are heavily based on the analysis of the environment. Imaging sensors, accelerometers or laser scanners are used to implement human-like behaviour of our robotic applications. However, AI based applications are typically not as reliable as needed for industrial needs.
In our research we tackle the explainability of novel deep learning and machine learning models with a strong focus on robot vision and robot motion. We implement complex applications and unveil the explanatory factors learned by the backbone of the models. By identifying and carefully analysing those explanatory factors, we can implement reliable robot application needed by industrial partners.