Enhancing Robotic Perception through Synchronized Simulation and Physical Common-Sense Reasoning.

Area of Science:

• Robotics
• Artificial Intelligence
• Cognitive Architectures

Background:

• Robotic cognitive architectures require efficient working memory systems.
• Intuitive physics principles are crucial for advanced robotic perception and interaction.
• Current methods for implementing complex robotic functionalities often involve extensive coding.

Purpose of the Study:

• To investigate the hypothesis that integrating a physics simulator into a cognitive architecture streamlines robotic functionality implementation.
• To categorize functionalities enhanced by physics simulation in robotics.
• To demonstrate empirical findings from this integration.

Main Methods:

• Amalgamation of the CORTEX cognitive architecture with an embedded physics simulator (CoppeliaSim).
• Utilizing deep state representation for the working memory, including ontology, state persistency, and relationships.
• Real-world experiments with Kinova Gen3 and Open-Manipulator-P robotic arms, synchronized with the simulator.

Main Results:

• Streamlined implementation of functionalities such as occlusion handling, model-based perception, self-calibration, scene structural stability, and human activity interpretation.
• Demonstrated enhancement of overall robotic agent performance through projected simulation results into working memory.
• Empirical validation of the benefits of physics simulation within a cognitive architecture.

Conclusions:

• The integration of physics simulators significantly simplifies the development of advanced robotic capabilities.
• This approach offers a more efficient pathway for robots to understand and interact with complex environments.
• The CORTEX architecture with embedded simulation provides a robust framework for future robotic research and development.