A Foundation for Perception in Autonomous Systems

Tesis Doctoral
Universidad Politécnica de Madrid

Author:

Ignacio López

Tutor:

Ricardo Sanz

Abstract

Perception is a phenomenon which takes occurs as part of the operation of some systems. The most complex perceptive systems are found in some biological creatures. Certain indefinition of the notion is due to divergencies between the different particular meanings of the term. Nevertheless, all views share, in a variably implicit form, that perception stands for a link between that which is perceived and that which perceives. Depending on the particular context considered, this link results in a modelization of the environment, a reactive acion, or a representation.

In these terms, the context in which perception takes place, in most general of cases, is an autonomous system. Although strictly autonomy as a concept is exclusive, we will admit degrees of autonomy in systems.

In a conceptual sense, perception will occur within the system as the consequence of the action of a Perceptor. It will generate Percepts.These are the result of a process called, like the phenomenon, Perception. Theset of all agents necessary to produce a percept within a given system will becalled Perceptive System; its components, configuration, capacity and all other characteristics are strictly defined within a particular system and instant of time.

As a consequence of the previous aspects mentioned, the perceptive phenomenon is structurally and operationally dependent on the rest of the autonomous system; this will be reflected on the corresponding percepts. Reciprocally, perception also conditions the rest of the system. The two main points on which this occurs are perception of time (thus determining to a great extent the operation of the system) and information processing.

This work represents a generalization of current studies onsensation and perception to the scope of general systems, and specifically addresses the clarification and formulation of a epistemological theory of naturaland artificial perception

Status of this work

The work has been structured in four parallel tasks:

General principles have been established; the construction of the framework is in progress. Phases 3 and 4 are in progress.

Scheduled defense

Thesis Objectives

The main objective is to build a general theory on perception. It will establish a consistent conceptual background on the matter, representing the different parts involved in perception within an autonomous system, as well as operational aspects. All will be treated at a high level of abstraction.

The theoretical corpus will serve as a basis for analyzing current perceptive systems in virtually any context, biological or artificial indistinctively. An indirect objective of the work is to address the problem of engineering perception at a high level of abstraction, by formulating a series of considerations in form of principles, concepts and notions derived from the general framework. In more practical terms, this work will cover advanced sensing, information processing and representation, as part of the problem of control in uncertain environments.

Potential Applications

The sensing process is relatively well understood and addressed by the theory of measure at the lower layers of a control system. In highlevel controllers, however, the sensing and perception process is strongly imbricated into the process of action generation. While this is efficient -and this the the evolutionary reason for its existence- it is, however, ineffective as an engineering paradigm for highly evolvable, adaptive architectures, designed for operation under uncertain environments.

Industry: Perceptive systems for plant integration and industrial control systems. The requirements are (1) capacity for integrating existing highly heterogeneous systems. (2) Controlling large plants under strict functioning specifications.

Artificial Intelligence: Information processing systems capable of analyzing and interpreting complex sensory data within the context of an autonomous system.

Documents and links

The SOUL Architecture

The Design of SOUL. A Blueprint for a Mind. Ricardo Sanz, et al..This document describes the SOUL cognitive control architecture basics.

The Integrated Control Architecture

The Integrated Control Architecture. Middleware for Complex Control Systems. Ricardo Sanz.This document describes the ICa integrated control architecture platform.

Selected References