Automated Coordination Strategy Design Using Genetic Programming for Dynamic Multipoint Dynamic Aggregation.

Area of Science:

• Robotics and Artificial Intelligence
• Operations Research
• Optimization

Background:

• Multirobot systems are crucial for tasks like bush fire elimination.
• The dynamic multipoint dynamic aggregation (MPDA) problem involves collaborative task completion with continuously arriving tasks.
• Existing methods struggle with the complexity of heterogeneous robots, dynamic task arrivals, and inter-robot/task relationships.

Purpose of the Study:

• To propose a novel model for the dynamic MPDA problem that accounts for its inherent complexities.
• To develop an efficient computational method for solving this dynamic optimization problem.
• To generate real-time, adaptive coordination strategies for heterogeneous multirobot systems.

Main Methods:

• A new genetic programming hyperheuristic (GPHH) was developed.
• Reactive coordination strategies (RCSs) were evolved using the GPHH.
• A novel RCS heuristic template, terminal set, and cluster filter were designed to guide robot decision-making.

Main Results:

• The proposed GPHH method significantly outperformed existing state-of-the-art approaches.
• The GPHH effectively evolved reactive coordination strategies for real-time robot decision-making.
• Experimental analysis provided insights into optimal robot distribution and coordination for diverse tasks.

Conclusions:

• The developed GPHH offers a superior solution for the dynamic MPDA problem in multirobot systems.
• The evolved RCSs enable robots to adaptively manage unpredictable task arrivals and collaborate effectively.
• This research advances the field of multirobot coordination, particularly for dynamic and complex operational environments.