Application of Event Cameras and Neuromorphic Computing to VSLAM: A Survey

Sangay Tenzin ¹, Alexander Rassau ¹, Douglas Chai ¹

⁰School of Engineering, Edith Cowan University, Perth, WA 6027, Australia.

Biomimetics (basel, Switzerland) +

|

July 26, 2024

View abstract on PubMed

Summary

Event cameras and neuromorphic processors enhance Simultaneous Localization and Mapping (SLAM) systems, overcoming limitations of traditional methods in challenging environments. This integration offers improved energy efficiency and real-time performance for autonomous navigation.

Area Of Science

Robotics and Autonomous Systems
Computer Vision
Neuromorphic Engineering

Background

Traditional Visual SLAM (VSLAM) faces limitations in dynamic or low-light conditions due to frame-based cameras and processing pipelines.
Event cameras offer asynchronous data capture with high temporal resolution and low power consumption, inspired by biological vision.
Neuromorphic processors provide brain-like parallel processing for efficient real-time analysis of event-based data.

Purpose Of The Study

To provide a comprehensive overview of integrating event cameras and neuromorphic processors into VSLAM systems.
To highlight the advantages of event-based sensing and neuromorphic processing over traditional approaches.
To survey state-of-the-art event-based SLAM techniques and their synergistic benefits.

Main Methods

Review of principles behind event cameras and neuromorphic processors.
In-depth survey of current event-based SLAM approaches: feature extraction, motion estimation, map reconstruction.
Exploration of the integration of event cameras with neuromorphic processors, analyzing benefits in energy efficiency, robustness, and real-time performance.

Main Results

Event cameras and neuromorphic processors significantly improve VSLAM performance in challenging environments.
Synergistic integration leads to enhanced energy efficiency, robustness, and real-time capabilities.
Identified key techniques in feature extraction, motion estimation, and map reconstruction for event-based SLAM.

Conclusions

Event-based SLAM, powered by event cameras and neuromorphic processors, represents a significant advancement over traditional VSLAM.
This technology holds immense potential for applications in robotics, autonomous vehicles, and augmented reality.
Further research is needed to address challenges in sensor calibration, data fusion, and algorithmic development.

Keywords:

Simultaneous Localization and Mapping (SLAM) autonomous systems event camera machine vision neuromorphic processing real-time processing robotics sensor fusion