Engineering 2 1156 High Street, Santa Cruz, California 95064

In environments where Global Navigation Satellite Systems (GNSS) are unavailable traditional navigation systems that rely on GNSS signals face significant challenges. Simultaneous Localization and Mapping (SLAM) offers a viable alternative, enabling autonomous robots to navigate and map their surroundings independently. However, reflective surfaces like mirrors pose substantial difficulties for SLAM systems by generating map artifacts and erroneous landmarks that distort environmental maps and hinder navigation.

This research proposes a novel approach to enhance SLAM by enabling autonomous robots to detect mirrors through robotic self-recognition. Utilizing a re-trained object detection model specifically adapted to recognize the robot's own reflection, the system identifies and classifies mirrors within the environment. This classification prevents mapping errors by distinguishing mirrors from genuine navigable spaces and landmarks. Additionally, mirrors are leveraged as sources of real-time feedback on the vehicle's pose, as the reflection provides additional information about its position and orientation.

By integrating self-recognition and mirror detection into the SLAM framework, this method not only filters out misleading data but also utilizes reflections to enhance localization accuracy. This innovative approach improves the accuracy and robustness of SLAM systems, leading to enhanced autonomous navigation performance in complex and dynamic GNSS-denied environments.

 

Event Host: Carlos Espinosa, Ph.D. Student, Computer Engineering

Advisor: Gabriel Elkaim

Event Details

See Who Is Interested

0 people are interested in this event


User Activity

No recent activity