Force Observer-Based Motion Adaptation and Adaptive Neural Control for Robots in Contact With Unknown Environments

Area of Science:

• Robotics
• Control Systems
• Artificial Intelligence

Background:

• Robots operating in unknown environments require robust control strategies.
• Sensing-only approaches for force estimation limit adaptability.
• Existing trajectory iteration methods have limitations in motion speed flexibility.

Purpose of the Study:

• To propose a spatial learning control system for robots in unknown environments.
• To enable robots to achieve desired behaviors through adaptive interaction.
• To overcome limitations of existing control methods regarding motion speed.

Main Methods:

• A force observer estimates interaction forces without dedicated sensing devices.
• A learning law updates the robot's reference trajectory to control interaction forces.
• An adaptive controller with neural networks compensates for system uncertainties.
• Lyapunov's theory is used to prove system stability and bounded states.

Main Results:

• The proposed system successfully maintains desired interaction forces.
• The method removes the fixed motion speed limitation of trajectory iteration algorithms.
• Neural network-based adaptive control ensures high control accuracy.
• Simulations and experiments confirm the system's effectiveness on a robot platform.

Conclusions:

• The spatial learning control system offers enhanced feasibility and adaptability for robots in unknown environments.
• The adaptive controller effectively compensates for system uncertainties, ensuring stability and accuracy.
• This approach represents a significant advancement in robot interaction control.