Online Nash Solution in Networked Multirobot Formation Using Stochastic Near-Optimal Control Under Dynamic Events.

Area of Science:

• Robotics
• Control Systems
• Artificial Intelligence

Background:

• Networked multirobot systems face challenges from uncertainties like packet loss and transmission delays.
• Multirobot formation control is often modeled as a nonzero-sum game.
• Existing control methods may not efficiently handle stochasticity and real-time demands.

Purpose of the Study:

• To propose an online stochastic dynamic event-based near-optimal controller for networked multirobot formation.
• To ensure system stability and achieve a Nash equilibrium for coordinated control.
• To minimize control updates through an event-based triggering mechanism.

Main Methods:

• Developed an event-based control strategy for networked multirobot systems.
• Utilized an actor-critic neural network architecture for online policy generation and weight updates.
• Derived theoretical guarantees for the ultimate boundedness of estimation and formation errors.

Main Results:

• The proposed controller achieves near-optimal control inputs and attains a Nash equilibrium.
• Online policy generation occurs only at event instants, reducing computational load.
• Real-time validation with Pioneer P3-Dx robots demonstrated the approach's efficacy.
• Control update instants were significantly minimized for follower robots (20% and 18%).

Conclusions:

• The event-based, actor-critic approach effectively manages network uncertainties in multirobot formations.
• The method ensures system stability and achieves desired formations with enhanced efficiency.
• This work offers a practical solution for robust and optimized control in networked robotic systems.