A Bio-Inspired Hybrid Slam Algorithm for Minimal-Sensor Autonomous Robotics

Abstract

This paper presents an insect and mammal inspired paradigm of Simultaneous Localization and Mapping (SLAM)
algorithm for minimal power robotics which preserves sensor autonomy features. It combines visual odometry with
lightweight LiDar, mapping, proprioceptive, and volumetric inputs in a modular structure that optimally utilizes
processing resources without loss in fidelity. We demonstrated one of the strategies by coupling a graph-based SLAM
with an insect-like visual homing behavior so that it could function in harsh contexts where GPS or sensor access was
denied. Also attention based models are incorporated for feature selection aiding data association to enhance reliability
for loop closure. The model integrates metrical and topological mapping dynamically switching based on
environmental complexity. This provides guarantee for well terrain agnostic performance throughout different
landscapes. Relative to other minimal sensory SLAM systems, we achieved outstanding localization accuracy,
enhanced map consistency, and verification in both simulations and on a micro robotic platform. The proposed
algorithms shift autonomous energy efficiency, advantageous for small graspable robots operating in congested or
confined spatial regions. The goal is to advance the research and development of autonomous robots with low
bioinspired power features and realistic robust navigation systems.