A Sampling-Based Bayesian Approach for Cooperative Multiagent Online Search With Resource Constraints.

Area of Science:

• Robotics
• Artificial Intelligence
• Optimization Algorithms

Background:

• Cooperative multiagent search is crucial for complex tasks.
• Existing methods often struggle with resource constraints and balanced consumption.
• Bayesian frameworks are effective for probabilistic target localization.

Purpose of the Study:

• To develop a cooperative multiagent search algorithm for efficient target detection on a 2-D plane.
• To address multiple constraints including resource consumption and balanced agent effort.
• To improve upon existing search algorithms in terms of efficiency and resource management.

Main Methods:

• Utilized a Bayesian framework for updating local probability density functions (PDFs) based on agent observations.
• Employed a sampling-based logarithmic opinion pool for fusing local PDFs into a global PDF.
• Applied Gaussian Mixture Models (GMM) and a weighted expectation maximization algorithm for PDF reconstitution and parameter estimation.
• Formulated a utility function-based optimization problem solved via a gradient-based approach to guide agents.

Main Results:

• The proposed algorithm demonstrated reduced overall resource consumption compared to existing methods.
• Achieved superior performance in balancing resource consumption among agents.
• Effectively integrated Bayesian updating, PDF fusion, and optimization for enhanced search efficiency.

Conclusions:

• The developed cooperative multiagent search algorithm offers significant advantages in resource efficiency and balanced consumption.
• The integration of advanced probabilistic methods and optimization techniques provides a robust solution for constrained search problems.
• This approach has the potential to enhance the performance of autonomous systems in various search and rescue or exploration applications.