XBot2D: towards a robotics hybrid cloud architecture for field robotics

Area of Science:

• Robotics
• Cloud Computing
• Artificial Intelligence

Background:

• Robots face challenges in unstructured environments, needing to balance computational performance with power autonomy.
• Cloud robotics offers a solution by offloading computation, but network constraints hinder real-world deployment.

Purpose of the Study:

• To design a software architecture for hybrid cloud robotics that addresses dynamic environments and network limitations.
• To enable seamless task allocation between local robot resources and cloud computing.

Main Methods:

• Developed the XBot2D hybrid cloud manager for dynamic resource allocation.
• Implemented a two-dimensional (2D) framework leveraging local and cloud resources.
• Supported Real-Time (RT) skills locally and AI/ML in the cloud, with automatic relocation based on network conditions.

Main Results:

• XBot2D dynamically allocates robotics skills for distributed computation.
• The framework transparently manages local and cloud resources based on network quality and robot capabilities.
• Validated with the CENTAURO robot and AWS EC2 under various network conditions.

Conclusions:

• XBot2D provides a robust solution for cloud-enabled field robotics.
• The architecture enhances robot adaptability, performance, and power autonomy in challenging scenarios.