Segmentation-Based Dynamic Path Planning Using Depth Camera for Navigation by a Companion Robot

This research addresses the need for improved vision-based techniques in indoor robot navigation, benefiting fields like self-driving cars and assistive technologies in medical robotics. The study emphasizes the significance of identifying object positions and predicting safe navigation paths by treating scene perception as a general problem. The proposed approach combines deep learning and basic calculations to effectively understand scenes and predict paths. The research demonstrates the method's efficacy in accurately estimating person positions and achieving safe robot navigation. These findings contribute to the advancement of robot navigation, making it safer, more reliable, and applicable in diverse real-world scenarios, benefitting society.