An Embedded MAS (Multi-Agent Systems) can use different communication infrastructures, exchange information, and even transport internal agents to another system.
These systems can interact with a dynamic environment through their resources (sensors and actuators).
However, resources are static and defined at design time.
The addition of a resource, no matter how simple, forces the shutdown of the embedded system and several changes are necessary for the layers of hardware, firmware, interfacing, and reasoning.
This work presents an approach that guarantees the expected adaptive capacity of a Cognitive MAS by adding resources in runtime at an Embedded SMA.