A Virtual Sensor for Collision Detection and Distinction with Conventional Industrial Robots

Area of Science:

• Robotics
• Human-Robot Interaction
• Control Systems

Background:

• Physical contact is unavoidable during robot operations.
• Distinguishing between safe contact and dangerous collisions is crucial for human safety and effective human-robot interaction.
• Existing methods may require complex calculations or additional sensors.

Purpose of the Study:

• To propose a novel virtual sensor for detecting and distinguishing contact from collision in human-robot interactions.
• To enhance the safety and performance of industrial robots during close interactions with humans.
• To develop a system that relies solely on readily available robot joint data.

Main Methods:

• Designed a virtual sensor based on the robot's generalized momentum.
• Implemented two observers with low-pass and band-pass filter characteristics for collision detection.
• Utilized the distinct frequency signatures of contact and collision forces for differentiation.
• Selected appropriate filter parameters based on force signal frequency distributions.
• Required only motor driving current and joint position data, avoiding acceleration or inverse dynamics calculations.

Main Results:

• The low-pass and band-pass torque observers successfully detected different force signals in real-time.
• The proposed virtual sensor effectively distinguished between contact and collision events.
• Experimental validation confirmed the system's capability in human-robot interaction scenarios.
• The system demonstrated applicability to conventional industrial robots.

Conclusions:

• The developed virtual sensor provides a reliable method for real-time collision detection and distinction in human-robot interactions.
• The system enhances safety by accurately identifying hazardous collisions versus incidental contact.
• Its reliance on basic joint data makes it easily integrable into existing industrial robot systems.