A Machine Learning Approach to Object Detection for Free Path Navigation

The autonomous navigation system poses a significant challenge to free path planning in real-time detection of moving objects, especially in unstructured environments where generalised methods fail to achieve desired outcomes. This paper presents an advanced Machine Learning technique, which will be implemented in object detection to further improve free path planning by autonomous systems. Herein we explicitly consider convolutional neural networks, which refers to deep learning algorithms for dynamic object detection in real-time and reinforcement learning for adaptive path planning. In turn, these approaches are merged into one framework that allows autonomous agents to effectively detect and track obstacles so that flexible but safe navigation can be achieved even in unpredictable environments. An analysis of the performance of various Machine Learning models in object detection tasks and their impact on path planning algorithms is presented. Results proven that object detection based on Machine Learning is significantly enhancing features of path planning in autonomous systems to be adaptive, safe and efficient. Therefore, this paper presents further steps toward improving competences in autonomous systems by proposing smarter navigation solutions for self-driving cars, robots and Unmanned Aerial Vehicle (UAV).