Multi-Agent Network Control: A Brain Emotional Learning-Inspired Approach
Abstract
The research objective of this project is to employ computational models of emotional learning observed in the mammalian limbic system to develop novel and systematic methodologies for analysis, design, and implementation of autonomous multi-agent systems (MAS) operations. The motivation comes from the interdisciplinary and complex nature of the tasks encountered in the modern society, which demand the integration of multiple complementary agents capable of self-organization and coordinate themselves. This work builds on the PIs previous results demonstrating that the implementation of Brain Emotional Learning (BEL) is effective in stabilizing a single autonomous aerial agent, as well as a team of robots, both in simulations and real-time experiments. Coordination of MAS in real-time missions is challenging because the dynamics of the robotic agents, which could be aerial, ground, water vehicles, or even a combination of them, are usually not precisely known.Furthermore, MAS operations are often subjected to external disturbances and varying operational conditions. It is hypothesized that BEL strategies will provide MAS with learning capabilities, multi-objective properties, and low computational complexity. To achieve these goals, the proposed project focuses on the following fundamental research thrusts: 1. Adaptation of reward and sensory signals: BEL-inspired controllers must be provided with sensory signals and emotional cues, which should make sense with respect to the MAS states and objectives. It is hypothesized that, by adaptively choosing these functions by means of self-organizing methods and multi-objective goals (e.g., flocking, obstacle avoidance, energy savings) the mission will be optimally performed. 2. Stability of BEL controllers for MAS: To enhance the theoretical contribution, stability conditions for BEL-inspired feedback controllers for MAS will be investigated. At present, there is not a specific way for ensuring stability of these methods.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 31, 2020
- Accession Number
- AD1113056
Entities
People
- Luis Rodolfo García Carrillo
Organizations
- Texas A&M University