[Paper Review] Technical Report: Artificial DNA - a Concept for Self-Building Embedded Systems

Uwe Brinkschulte|arXiv (Cornell University)|Jul 24, 2017

Modular Robots and Swarm Intelligence12 references3 citations

TL;DR

This paper proposes Artificial DNA—a self-describing, self-building blueprint for embedded systems stored in every processor core—enabling runtime self-organization, self-repair, and dynamic adaptation. By encoding system structure, task relationships, and hardware suitability in a compact digital DNA, the system autonomously configures and reconfigures itself using a decentralized Artificial Hormone System (AHS), achieving real-time, fault-tolerant operation in complex, large-scale embedded systems.

ABSTRACT

This technical report deals with the concept of an artificial DNA which contains a blueprint of the structure and organization of an embedded system. This blueprint can be used to build up the embedded system in a self-organizing manner at run-time. The report describes in detail the basic principles of the the artificial DNA and its relationship to standard design methods for embedded systems. A prototypic implementation is presented and evaluated. Additionally, future work is described and a conclusion is given.

Motivation & Objective

• To address the growing complexity of embedded systems with thousands of computing nodes, which manual design cannot scale to.
• To enable self-organization, self-healing, and self-optimization in distributed real-time embedded systems.
• To develop a self-describing system architecture inspired by biological DNA, allowing runtime system construction and adaptation.
• To support dynamic fault tolerance through runtime replication and redundancy synthesis based on importance and resource availability.
• To enable evolution and reconfiguration of system structure at runtime using the artificial DNA as a dynamic blueprint.

Proposed method

• Represent embedded systems as networks of basic functional elements, each identified by a unique ID and parameter set, modeling behavior via state and output functions.
• Encode the complete system structure, task dependencies, communication, and core suitability into a compact digital 'Artificial DNA' stored in every processor core.
• Use the Artificial Hormone System (AHS) to transcribe the Artificial DNA at runtime, dynamically assigning tasks to cores based on current system state and constraints.
• Implement modular redundancy by dynamically creating replicas of critical basic elements based on importance values and resource availability, enabling self-adaptive fault tolerance.
• Support dynamic reconfiguration and evolution by allowing the DNA to be modified at runtime, with replicas and system structure adapting automatically to changes.
• Enable implicit redundancy by having multiple instances of the same basic element (e.g., ALUs, PID controllers) perform mutual validation through challenge-response checks.

Experimental results

Research questions

• RQ1 How can a self-describing, self-building system be designed for large-scale embedded systems with thousands of cores using bio-inspired principles?
• RQ2 What is the optimal way to represent system structure and behavior in a compact, distributed, and evolvable digital DNA?
• RQ3 How can modular redundancy be synthesized dynamically at runtime based on importance and resource constraints?
• RQ4 How can self-healing and fault tolerance be achieved through implicit and explicit replication of basic elements in the artificial DNA?
• RQ5 To what extent can the artificial DNA support real-time system reconfiguration and evolution without manual intervention?

Key findings

• The prototypical implementation of the Artificial DNA approach was successfully evaluated on a real-world robot vehicle, demonstrating feasibility and practicality.
• The memory and communication overhead of the Artificial DNA system is low, with compact DNA representations achievable using a limited set of basic elements.
• Self-building, self-adapting, and self-repairing behaviors based on the DNA blueprint can meet real-time system requirements.
• Dynamic modular redundancy can be synthesized at runtime, with replicas relocating, adding, or removing based on resource availability and system needs.
• Implicit redundancy through multiple instances of the same basic element allows for low-overhead mutual validation and failure detection.
• The system supports automatic adaptation to changes in DNA structure, enabling self-healing and evolution without manual reconfiguration.