Cognitive decisions based on a rule-based fuzzy system

Abstract

We develop an agent-based artificial general intelligent system that can be implemented in compact and power-efficient electronic hardware. The hardware under development is called the Street Engine, which is a hardware-based cognitive architecture for implementing agent-based artificial intelligence. In this paper, we introduce an agent-based system to replicate simple cognitive behaviours. In the processes of this system, numerical data are converted into fuzzy symbolic representations of the surrounding environment, and reasoning rules are included in a modified Fuzzy Inference System to support the cognitive decision-making. We use a case study example, the homing behaviour of the honey bee, to demonstrate constructing production rules and implementing the cognitive and reasoning capabilities of agents. The low level cognitive behaviour is converted into a rule-based fuzzy system, and hardware-based experiments have been conducted to verify the effectiveness of the proposed technique.

Introduction

For more than half a century, Artificial Intelligence (AI) research has made a great impact on a wide range of industries. Particularly in the recent decade, machine-learning (ML) approaches [1], [43], [5] have been successfully applied in practical commercial applications. In ML approaches, knowledge and information are represented as numerical data [23]. ML uses feature extraction and classification techniques to recognise patterns in the data and use these as the basis for classification or prediction future data inputs. This technique has been applied with great success in applications such as imaging processing [24], [13] and big data analysis [28].

The “black box” processing of ML creates opaque algorithms and does not work so well in applications in which the input data cannot be characterised by frequently occurring patterns, for example when the machine enters an unfamiliar environment. In these circumstances, a rule-based approach such as implemented in a cognitive architecture shows advantages [22], [14], [48]. However, the principal difficulty with cognitive agents is that, as more advanced features are required, the amount of data acquired and the lifetime of the agents increase. The size of the real-time working memory increases rapidly with more advanced agents. Consequently, the lifetime and complexity of the agents are limited by the memory capacities and the performance of platforms on which they are run. To increase performance, the only practical options are to increase memory capacity or processor performance or more likely both. However, there are practical limits to what can be achieved this way, and the capacity of current agents falls short of human-level intelligence. Therefore, we are considering a new approach to the implementation of agents, with new processing methods and data structures, which are more amenable to implementation and scaling in microelectronic hardware.

We have devised a new type of computational processor with low power consumption, which we call the Street Engine [11]. This new processor is intended to implement a cognitive architecture and to be used for cognitive agents in autonomous devices. The agent is coded in a customised rule-based language based on OPS-5 [10], called the Street Language [11], [33], [47].

To demonstrate the feasibility of the Street Language/Street Engine approach to the implementation of cognitive agents, we have sought to replicate a simple form of cognitive behaviour found in nature. Neuroethological research [30] and training experiments of insects [15], [35], [7], [45] show that cognitive behaviours are the integration of serial multiple natural reactions, which are lower levels of stimuli triggered by a part of a hard-wired neural system. Even with a simple neural system such as found in insects, advanced structures in their neural networks integrate the functions of different sections together to achieve a range of cognitive behaviours. We choose to build a Street agent-based system to replicate the homing behaviour of honey bees and use this project to demonstrate that it is feasible to build agents with simple cognitive behaviours using the Street rule-based cognitive architecture.

In this paper, Section 2 introduces basic principles and general logic of the honey bee cognitive behaviours, and Section 3 presents our novel processing architecture. The details of the implemented system architecture, data sensing and representation and cognitive logic are given in Section 4. To show the performance of the proposed new approach, the system behaviours are analysed and evaluated in Section 5.