Towards a BDI Architecture for Cooperative Agents on Resource-Constrained Microcontrollers

Abstract

This paper presents the design and implementation of cooperative multi-agent systems based on the BDI (Belief-Desire-Intention) architecture for resource-constrained microcontrollers. The proposed solution is validated on real hardware, demonstrating a complete and decentralized cooperation cycle over a low-power ESP-NOW network. At the core of our architecture is a belief synchronization loop, a mechanism that integrates network-state awareness into the BDI reasoning cycle. This enables a sensor agent to deliberate on its environment and successfully request actions from a remote actuator agent by forming dynamic cooperation plans. This work aims to contribute to an identified gap in the literature concerning real world applications of cooperative agents on low-cost hardware, as research has predominantly focused on simulation, single-agent optimization, or higher-power computing platforms.