This thesis addresses the smart control methods for industrial robots, their applications, and learning strategies, in particular focusing on efficient data preparation, computer vision-based robot control, and knowledge acquisition from humans through demonstrations. The primary aim of the Thesis is to research and develop novel techniques in these fields to improve the potential deployment of industrial robots for working in dynamic environments and advance the operability of such systems. Based on the identified challenges and open issues during the literature review, the Thesis develops and evaluates novel synthetic data generation techniques for usage in computer vision-based robot control and an approach for learning from human demonstrations.