Cognitive decisions based on a rule-based fuzzy system
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.
Section snippets
Honey bee behaviour
Honey bees are believed to be one of the few types of insects that are capable of some low-level cognition [30], [35]. The research of Fry et al. [12], Menzel et al. [32] and Labhart et al. [21] has shown that honey bees exhibit cognitive behaviours including homing behaviour using landmark references and exchanging knowledge of the paths to the food sources through what is known as the “waggle dance” during which honey bees move with special patterns to explain their experiences on the path to
The street engine
Our group is currently developing a new type of computer processor based on a parallel processing rule-based system. It is designed so that it can be implemented as a power-efficient processor to be embedded in autonomous hardware. The language that we use to express system behaviour is called the Street Language and the hardware architecture that supports execution of this language is called the Street Engine [11].
System implementation
One main objective of this work is to demonstrate the feasibility and capacity of the proposed approach through the implementation of a case study. Therefore, we chose to implement a Street agent-based system on a two-wheel drive rover platform, and test it in an artificial environment with coloured balls as landmarks. The system is designed to reproduce the honey bee’s homing behaviour, and to imitate the way a honey bee processes. In this simulation, electronic sensors are used to replicate a
Foraging mode
The system starts in foraging mode, in which the Command Agent instructs the platform to explore the surrounding environment by generating a series of pseudo-random movements. As it does so, the Memory Agent determines significant changes in the surrounding environment to recognise new reference positions and records the sequence of these reference positions and their characteristics, using MFs.
A _INPUT MF is pushed into the WMs of both agents in real-time, containing a set of _INPUT WMEs
Conclusion
This paper has detailed the conceptual design, development and testing of a system which is based on rule-based processing using fuzzy representation of the environment to reproduce honey bee homing behaviours. The system is able to approximately reproduce the homing behaviour of honey bees in an artificial environment. The same method and system can be scaled up and used for duplicating human cognitive behaviours. This work has demonstrated the feasibility of using a rule-based approach for
CRediT authorship contribution statement
Xin Yuan: Investigation, Methodology, Software, Writing – original draft. Michael John Liebelt: Supervision, Writing – review & editing. Peng Shi: Supervision, Writing – review & editing. Braden J. Phillips: Conceptualization.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgement
aaa
References (49)
- et al.
A generalized type-2 fuzzy logic approach for dynamic parameter adaptation in bee colony optimization applied to fuzzy controller design
Inf. Sci.
(2018) - et al.
A generalized type-2 fuzzy granular approach with applications to aerospace
Inf. Sci.
(2016) - et al.
Transfer learning based countermeasure against label flipping poisoning attack
Inf. Sci.
(2021) - et al.
Neural fuzzy inference systems with knowledge-based cultural differential evolution for nonlinear system control
Inf. Sci.
(2014) - et al.
The dance of the cape gannet may contain social information on foraging behaviour
Anim. Behav.
(2020) - et al.
Earc: Evidential association rule-based classification
Inf. Sci.
(2021) - et al.
Local optimality of self-organising neuro-fuzzy inference systems
Inf. Sci.
(2019) Some labels that are recognized on landmarks by the honeybee (apis mellifera)
J. Insect Physiol.
(2006)- et al.
Neural mechanisms in insect navigation: polarization compass and odometer
Curr. Opin. Neurobiol.
(2002) - et al.
Further robust stability analysis for uncertain takagi-sugeno fuzzy systems with time-varying delay via relaxed integral inequality
Inf. Sci.
(2017)
A fuzzy envelope for hesitant fuzzy linguistic term set and its application to multicriteria decision making
Inf. Sci.
Cognitive architecture of a mini-brain: the honeybee
Trends Cogn. Sci.
A common frame of reference for learned and communicated vectors in honeybee navigation
Curr. Biol.
Finite-interval-valued type-2 gaussian fuzzy numbers applied to fuzzy todim in a healthcare problem
Eng. Appl. Artif. Intell.
Data-driven dynamic resource scheduling for network slicing: A deep reinforcement learning approach
Inf. Sci.
Dissipativity analysis and synthesis for positive roesser systems under the switched mechanism and takagi-sugeno fuzzy rules
Inf. Sci.
Fuzzy sets
Inf. Control
Introduction to machine learning
The fine structure of honeybee head and body yaw movements in a homing task
Proc. R. Soc. B
The spectral input to honeybee visual odometry
J. Exp. Biol.
Honey bee foraging distance depends on month and forage type
Apidologie
OPS5 user’s manual
A new computer for cognitive computing
Honey bees store landmarks in an egocentric frame of reference
J. Comp. Physiol. A
Cited by (3)
Information granulation-based fuzzy partition in decision tree induction
2022, Information SciencesCitation Excerpt :The forms of membership function in fuzzy items have been widely studied in many fuzzy systems. Various forms of membership functions have been suggested such as triangular form, trapezoidal form, Gaussian form and Z-shaped [38–40]. This study utilizes the mixed types membership functions to define fuzzy items, that is, the first fuzzy item uses the right trapezoidal form, the last fuzzy item utilizes the left trapezoidal form, and the rest fuzzy items take the triangular form.
Knowledge-Embedded Mutual Guidance for Visual Reasoning
2024, IEEE Transactions on CyberneticsFormation of Subsets of Co-expressed Gene Expression Profiles Based on Joint Use of Fuzzy Inference System, Statistical Criteria and Shannon Entropy
2023, Lecture Notes on Data Engineering and Communications Technologies