Adaptive stochastic gradient descent with least angle regression enhanced navigation: intelligent path planning in cluttered environments for autonomous robots

Abstract

In the dynamic realm of Autonomous Mobile Robots (AMRs), ensuring smooth navigation among obstacles is critical, especially as they become increasingly integral to industries such as manufacturing and transportation. Recent advances have introduced several learning models to aid in obstacle avoidance, but many face computational challenges. This research introduces the Adaptive Stochastic Gradient Descent with Least Angle Regression (ASGD-LARS) algorithm, specifically designed to enhance the navigation of AMRs. By carefully considering obstacle orientations, it facilitates quicker decision-making for direction changes. When compared with well-established algorithms like KNN, XG Boost, Naive Bayes, and Logistic Regression, ASGD-LARS consistently performs better in terms of accuracy, computational efficiency, and reliability. This study lays the foundation for the deployment of smarter and more efficient AMRs across diverse industries.