Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Building a Cognitive Twin Using a Distributed Cognitive System and an Evolution Strategy

Wandemberg Gibaut, Ricardo Gudwin

TL;DR

The paper tackles creating a Cognitive Twin that can emulate a user’s interaction behavior by leveraging a distributed cognitive framework (DCT) and an Evolution Strategy to optimize inter-codelet topology. It introduces a four-codelet agent (Sensory, Perceptual, Behavioral, Motor) implemented on virtual devices and trained offline to map sensor inputs to actuator outputs, followed by end-to-end training within a distributed environment. Experimental results in a smart-home-like scenario show that a substantial fraction of runs achieve low error with topology typically involving a moderate number of Behavioral Codelets, and that more Behavioral Codelets generally improve performance. The work demonstrates a scalable, low-power approach to cognitive twin construction with potential applications in automation, human-like agents, and deeper behavioral analysis.

Abstract

This work presents a technique to build interaction-based Cognitive Twins (a computational version of an external agent) using input-output training and an Evolution Strategy on top of a framework for distributed Cognitive Architectures. Here, we show that it's possible to orchestrate many simple physical and virtual devices to achieve good approximations of a person's interaction behavior by training the system in an end-to-end fashion and present performance metrics. The generated Cognitive Twin may later be used to automate tasks, generate more realistic human-like artificial agents or further investigate its behaviors.

Building a Cognitive Twin Using a Distributed Cognitive System and an Evolution Strategy

TL;DR

The paper tackles creating a Cognitive Twin that can emulate a user’s interaction behavior by leveraging a distributed cognitive framework (DCT) and an Evolution Strategy to optimize inter-codelet topology. It introduces a four-codelet agent (Sensory, Perceptual, Behavioral, Motor) implemented on virtual devices and trained offline to map sensor inputs to actuator outputs, followed by end-to-end training within a distributed environment. Experimental results in a smart-home-like scenario show that a substantial fraction of runs achieve low error with topology typically involving a moderate number of Behavioral Codelets, and that more Behavioral Codelets generally improve performance. The work demonstrates a scalable, low-power approach to cognitive twin construction with potential applications in automation, human-like agents, and deeper behavioral analysis.

Abstract

This work presents a technique to build interaction-based Cognitive Twins (a computational version of an external agent) using input-output training and an Evolution Strategy on top of a framework for distributed Cognitive Architectures. Here, we show that it's possible to orchestrate many simple physical and virtual devices to achieve good approximations of a person's interaction behavior by training the system in an end-to-end fashion and present performance metrics. The generated Cognitive Twin may later be used to automate tasks, generate more realistic human-like artificial agents or further investigate its behaviors.

Paper Structure

This paper contains 23 sections, 9 figures, 1 table.

Table of Contents

  1. Introduction
  2. Cognitive Systems
  3. Digital and Cognitive Twins
  4. Systems of Systems
  5. Cyber-Physical Systems
  6. Related Work
  7. The DCT
  8. The DCT Architectural Overview
  9. The DCT Codelet structure
  10. The DCT Memory structure and default support
  11. DCT Node
  12. An Evolutionary Cognitive Twin
  13. Devices internal structure
  14. Sensory Codelets
  15. Perceptual Codelets
  16. ...and 8 more sections

Figures (9)

  • Figure 1: Illustration of a multi-device, Codelet-oriented system as seen in gibaut2020extending. Notice that, depending on how powerful is the device, it may run a single Codelet or multiple ones.
  • Figure 2: The concept of a DCT Codelet.
  • Figure 3: Graphical representation of an agent structure and its internal connections
  • Figure 4: Example Individual for our Evolution Strategy Process. The Individual is encoded as an array of binary values, each one representing if a certain Perceptual or Behavioral Codelet is to be considered as part of the agent.
  • Figure 5: Evolution Strategy process diagram. Here we have a high-level representation of each step of the mentioned heuristic.
  • ...and 4 more figures

Theorems & Definitions (1)

  • Definition 2.1: A DCT Cognitive System