Decoupling what, how, and when for observing decision-making context in autonomous robots

Area of Science:

• Robotics
• Artificial Intelligence
• Computer Science

Background:

• Situation awareness is crucial for autonomous robot reliability.
• Current systems struggle to integrate perception with real-time decision-making and action execution.
• Perception often lacks context-awareness, hindering efficient operation.

Purpose of the Study:

• To develop a novel approach for decoupling observation, perception, and decision-making in autonomous robots.
• To enhance situation awareness and operational efficiency in robotic systems.
• To improve the robustness and speed of autonomous robot task execution.

Main Methods:

• Explicitly modeling decision-making context using dedicated observers.
• Implementing observer behaviors as behavior trees coordinated by a centralized context manager.
• Parallelizing perception tasks with ongoing actions for real-time adaptation.

Main Results:

• Context-driven observers successfully identified decision-making contexts without disrupting ongoing actions.
• Parallelized perception led to significant runtime improvements.
• Achieved overall task time savings of up to 24% in machine-tending tasks.

Conclusions:

• Explicit context modeling and observer-based perception parallelization enhance robotic execution architectures.
• The proposed approach improves both the stability and efficiency of autonomous robots.
• This method offers a robust solution for real-world applications in complex environments.