A Self-Aware and Scalable Solution for Efficient Mobile-Cloud Hybrid Robotics.

Area of Science:

• Robotics
• Cloud Computing
• Optimization

Background:

• Mobile-cloud hybrid (MCH) robotic tasks face challenges in optimizing conflicting objectives like battery and network usage.
• Existing MCH approaches often lack efficiency in balancing these competing demands.

Purpose of the Study:

• To propose a novel approach for instrumenting MCH robotic tasks and searching for efficient configurations.
• To develop a framework for runtime measurement of MCH task objectives and a multi-objective optimization algorithm.

Main Methods:

• Introduced a general-purpose MCH framework for runtime measurement of battery consumption and network usage.
• Developed a novel two-step search-based multi-objective optimization (MOO) algorithm to find efficient MCH configurations.
• Implemented self-adaptive and self-aware decision-making based on environmental and network changes.

Main Results:

• MCH foraging tasks on battery-powered robots achieved better optimization with self-adaptive/self-aware decisions compared to static offloading or robot-only execution.
• Self-aware systems demonstrated superior performance in minimizing objectives during internal system changes.
• The Two-Step MOO algorithm effectively identified high-quality configurations for small to medium-scale MCH robotic tasks.

Conclusions:

• The proposed framework and algorithm enable efficient optimization of MCH robotic tasks in dynamic environments.
• Self-aware decision-making is crucial for robust performance when internal system changes occur.
• The developed MOO algorithm provides a scalable solution for configuring MCH robotic applications.