Qualia Phase-Space Prototype NEW: Control-Theoretic Modeling of Qualia as Safety Markers in Autonomous

Fuente: Zenodo
Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: Suvorov, Vladimir
Format: Recurso digital
Sprache:Englisch
Veröffentlicht: Zenodo 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866902205345300480
author Suvorov, Vladimir
author_facet Suvorov, Vladimir
contents <p class="MsoNormal"><span>This repository provides a prototype implementation and CARLA-based validation package supporting the study "Qualia as Control-Theoretic Constructs for Autonomous Agents: Event Phase Space as an Action-Oriented Semantic Safety Layer".</span></p> <p class="MsoNormal"><span>The archive contains source code, trained models, validation scripts, and output data used to model qualia-indicating signals as integrated control-relevant markers in an event-based phase space.</span></p> <p class="MsoNormal"><span>Within this framework, qualia are operationalized as pre-reflective, integrative assessment structures derived from the geometric proximity of the system's trajectory to critical regions (attractors, separatrices, saddle points) in a learned phase space. These structures function as safety markers and directly influence control policies in autonomous agents.</span></p> <p class="MsoNormal"><span>The package demonstrates:<br>- construction of an event phase space with learned vector field dynamics,<br>- trajectory tracking, attractor detection, and separatrix proximity estimation,<br>- generation of qualia-indicating alarm signals from unsupervised dynamics learning,<br>- validation via CARLA "Parked Vehicle Occlusion" scenarios (1,000 episodes per controller, 5 seeds, simple and stress variants),<br>- comparison of seven controllers including WM+ODE+alarm, WM-only, and conventional baselines (TTC, distance-reactive, phantom-agent).</span></p> <p class="MsoNormal"><span>Key results reproduced by this package: alarm AUC = 0.812, 29.6% collision-rate reduction in stress scenarios relative to a world-model-only baseline, 1.9 ms/decision on CPU with 6,550 parameters.</span></p> <p class="MsoNormal"><span>This artifact is intended for conceptual and empirical validation of the proposed control-theoretic interpretation of qualia and does not represent a production-ready system.</span></p> <p class="MsoNormal"><span>See README for structure, execution instructions, and data description.</span></p>
format Recurso digital
id zenodo_https___doi_org_10_5281_zenodo_19680882
institution Zenodo
language eng
publishDate 2026
publisher Zenodo
record_format zenodo
spellingShingle Qualia Phase-Space Prototype NEW: Control-Theoretic Modeling of Qualia as Safety Markers in Autonomous
Suvorov, Vladimir
qualia, event phase space, autonomous agents, world models, CARLA simulation, safety validation, embodied AI, control barrier functions, attractor dynamics, neural ODE
<p class="MsoNormal"><span>This repository provides a prototype implementation and CARLA-based validation package supporting the study "Qualia as Control-Theoretic Constructs for Autonomous Agents: Event Phase Space as an Action-Oriented Semantic Safety Layer".</span></p> <p class="MsoNormal"><span>The archive contains source code, trained models, validation scripts, and output data used to model qualia-indicating signals as integrated control-relevant markers in an event-based phase space.</span></p> <p class="MsoNormal"><span>Within this framework, qualia are operationalized as pre-reflective, integrative assessment structures derived from the geometric proximity of the system's trajectory to critical regions (attractors, separatrices, saddle points) in a learned phase space. These structures function as safety markers and directly influence control policies in autonomous agents.</span></p> <p class="MsoNormal"><span>The package demonstrates:<br>- construction of an event phase space with learned vector field dynamics,<br>- trajectory tracking, attractor detection, and separatrix proximity estimation,<br>- generation of qualia-indicating alarm signals from unsupervised dynamics learning,<br>- validation via CARLA "Parked Vehicle Occlusion" scenarios (1,000 episodes per controller, 5 seeds, simple and stress variants),<br>- comparison of seven controllers including WM+ODE+alarm, WM-only, and conventional baselines (TTC, distance-reactive, phantom-agent).</span></p> <p class="MsoNormal"><span>Key results reproduced by this package: alarm AUC = 0.812, 29.6% collision-rate reduction in stress scenarios relative to a world-model-only baseline, 1.9 ms/decision on CPU with 6,550 parameters.</span></p> <p class="MsoNormal"><span>This artifact is intended for conceptual and empirical validation of the proposed control-theoretic interpretation of qualia and does not represent a production-ready system.</span></p> <p class="MsoNormal"><span>See README for structure, execution instructions, and data description.</span></p>
title Qualia Phase-Space Prototype NEW: Control-Theoretic Modeling of Qualia as Safety Markers in Autonomous
topic qualia, event phase space, autonomous agents, world models, CARLA simulation, safety validation, embodied AI, control barrier functions, attractor dynamics, neural ODE
url https://doi.org/10.5281/zenodo.19680882