3D Recognition Based on Sensor Modalities for Robotic Systems: A Survey

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• 3D visual recognition is crucial for autonomous robots in real-world applications like object tracking and human-robot interaction.
• Social robots require robust environmental understanding for effective task assistance in professional and domestic settings.
• Recent advancements in deep learning have significantly impacted 3D recognition capabilities.

Purpose of the Study:

• To analyze data representation methods for 3D object and place recognition using deep learning.
• To highlight the value of social robots and their reliance on advanced recognition.
• To identify research gaps and provide a comprehensive overview of the field.

Main Methods:

• Analysis of data representation methods based on various sensor modalities.
• Application of deep learning models for 3D object and place recognition.
• Review and comparison of existing 3D recognition datasets and performance metrics.

Main Results:

• Identified key deep learning approaches for 3D visual recognition in robotics.
• Summarized current 3D recognition datasets and their performance benchmarks.
• Delineated critical research gaps and future research directions.

Conclusions:

• Recent developments in deep learning-based 3D visual recognition are foundational for advancing robotics.
• This survey serves as a guide for researchers and developers in vision-based robotics.
• Further research is needed to address identified gaps and enhance robot perception capabilities.