SmartWay introduces a zero-shot framework for Vision-and-Language Navigation that significantly improves how AI agents interpret natural language instructions while navigating complex 3D environments.

• Solves critical spatial awareness limitations in current waypoint predictors
• Implements sophisticated historical reasoning for better decision-making
• Incorporates intelligent backtracking capabilities to recover from navigation errors
• Designed for zero-shot performance without requiring environment-specific training

This research addresses fundamental engineering challenges in robotic navigation, enabling more reliable autonomous systems that can understand natural language commands in unfamiliar environments.

SmartWay: Enhanced Waypoint Prediction and Backtracking for Zero-Shot Vision-and-Language Navigation