Adjacent-Agent Dynamic Linearization-Based Iterative Learning Formation Control.

Area of Science:

• Control Theory
• Robotics
• Networked Systems

Background:

• Multiagent systems (MASs) present complex control challenges due to interconnected agent dynamics.
• Enhancing formation control performance in MASs is crucial for coordinated tasks.
• Existing methods often rely on detailed physical models, limiting applicability.

Purpose of the Study:

• To develop a novel data-driven control strategy for improving multiagent formation control.
• To address nonlinear and nonaffine dynamics in MASs using adjacent agent interactions.
• To propose a method that learns control policies from operational data.

Main Methods:

• Adjacent-agent dynamic linearization to create a virtual linear difference model between communicating agents.
• Assigning parent and child roles to agents to establish causal communication links.
• Adjacent-agent dynamic linearization-based iterative learning formation control (ADL-ILFC) for child agents, utilizing iterative learning from parent agents, time, and past actions.

Main Results:

• Demonstrated enhancement of formation control performance in MASs.
• Successfully applied a data-driven approach, eliminating the need for first-principle models.
• Validated the efficacy of ADL-ILFC through rigorous analysis and simulations.

Conclusions:

• The proposed ADL-ILFC method effectively improves multiagent formation control.
• The data-driven approach offers a flexible alternative to model-based control strategies for MASs.
• The method leverages iterative learning and agent interactions for robust control.