An IoT Autonomic Architecture for Self-replicating and Self-assembly Systems
Abstract
The number and diversity of Internet of Things (IoT) application domains are continuously in-creasing as the connected devices are doing exponentially. This has led to new challenges for the management, maintenance, and evolution of IoT systems. These issues have been approached by enabling adaptability and self-management capability (autonomic computing) in such systems.
However, the focus on this area has been more on the behavioral or functional side (algorithmic) of systems rather than on the architecture and evolution of the high-level system design. This paper
addresses this aspect by studying and proposing two new self-* properties for autonomic systems: Self-Assembly and Self-replication. These systems are referred to as extended autonomic systems.
The proposed architecture is based on a multi-agent approach, biologically inspired, and on the principle of maximum potential functionality: “installed hardware capacity, enabled software capac-
ity”, Although such behavior could lead to having source code that does not run on the devices (Passive Code) despite module beloging to the same community. The functional unit defined to perform self-assembly and self-replication of system functionality is located at the source code class level. The proposed architectural design and algorithms were validated through the implementation of “Quines” in Python, static code testing, and an agent-based simulation. The experimental test scenarios consist of cyber-physical hydroponic systems for urban agronomy which were simulated as agent communities. Results show that the number of simulation cycles required for code composition and distribution increases reasonably as a function of the agent’s community size and the 3 number of functional classes involved in Self-assembly and Self-replication. Finally, it is shown that an IoT system can replicate part of its functional design in another system that requires it.
However, the focus on this area has been more on the behavioral or functional side (algorithmic) of systems rather than on the architecture and evolution of the high-level system design. This paper
addresses this aspect by studying and proposing two new self-* properties for autonomic systems: Self-Assembly and Self-replication. These systems are referred to as extended autonomic systems.
The proposed architecture is based on a multi-agent approach, biologically inspired, and on the principle of maximum potential functionality: “installed hardware capacity, enabled software capac-
ity”, Although such behavior could lead to having source code that does not run on the devices (Passive Code) despite module beloging to the same community. The functional unit defined to perform self-assembly and self-replication of system functionality is located at the source code class level. The proposed architectural design and algorithms were validated through the implementation of “Quines” in Python, static code testing, and an agent-based simulation. The experimental test scenarios consist of cyber-physical hydroponic systems for urban agronomy which were simulated as agent communities. Results show that the number of simulation cycles required for code composition and distribution increases reasonably as a function of the agent’s community size and the 3 number of functional classes involved in Self-assembly and Self-replication. Finally, it is shown that an IoT system can replicate part of its functional design in another system that requires it.
Keywords
IoT, architecture, system