Navigating without GPS presents a major challenge for autonomous robots—especially with just a single onboard camera. This research introduces an innovative real-time navigation system that thrives in unknown, obstacle-filled environments.

Key breakthroughs include:

• Smart Mapping: A first-of-its-kind variable-resolution ESDF map that dynamically adjusts based on the robot’s speed, balancing detail and efficiency.

• Memory-Augmented Vision: A vicinity-based memory compensates for the camera’s limited view, enabling smoother obstacle avoidance.

• Lightweight Performance: The system runs without heavy computation, even outperforming state-of-the-art methods in complex dynamic tests.

Validated on real differential-drive robots, this approach proves that minimal sensors can deliver maximum autonomy—from warehouses to disaster zones.