1 of 21

The XI International Scientific and Practical Conference�"Information Control Systems &Technologies (ICST-Odessa-2023)"�����Cognitive Perception as a Base Model of the Feeling Artificial Intelligence��Anatolii Kargin and Tetyana Petrenko� kargin@kart.edu.ua petrenko_tg@kart.edu.ua ��Ukrainian State University of Railway Transport, Kharkiv, Ukraine���21 of September - 23 of September 2023�Odessa, Ukraine

2 of 21

� �Generative Artificial Intelligence (gen AI)�(OpenAI ChatGPT, Google Translate, IBM Watsonx, Apple Siri)�Artificial General Intelligence (AGI)�(DeepMind AlphaGo, AlphaZero, Thrill-K blueprint)��Feelling Artificial Intelligence (FAI)�(Martin Lockheed autonomous unmanned systems,�Google Waymo autonomous vehicles)

2

3 of 21

Autonomous vs. Automated vs. Self-Driving�The Society of Automotive Engineers (SAE) defines 6 levels of driving automation: from Level 0 (no automation) to Level 5 (full automation)��The SAE uses the term automated instead of autonomous. A autonomous car would be self-aware and capable of making its own choices��An Autonomous Intelligent Unmanned Systems (AIUS) is capable of feeling its environment and operating (perform its mission for any situation) without human involvement

3

4 of 21

Factors that violate the AIUS autonomy

Cause

Violation

Remedy

Large dimension of the decision-making space represented by a set of granules data from sensors

The solution cannot be obtained for some areas of this space

Reducing the decision-making space while preserving the meaning of the original sensor data set

Threshold boundaries of granulation of data from sensors.

Incorrect decision because real data from sensors not assigned to the corresponding granules due to corrupted inputs

Fuzzy representation and processing data granules

Knowledge incompleteness due to the complexity of the US environment, generated by a large number of objects and connections between them as also by dynamics of their changes.

A stage of the action plan cannot be completed due to low confidence in its feasibility in the current situation

Using the experience of implementing the plan in various similar situations.

Using an environmental exploration strategy to obtain missing information

Multi-stage plan can’t fully implement due to a predetermined planning strategy with strong restrictions on plan stages

Using the strategy of continuous conditional goal-setting planning with fuzzy restrictions on the stages of the plan.

Using the mechanism of the data meaning correction, taking into account the cognitive assessment of the situation

Different US components interpret and understand the meaning of the same data in different ways.

Solutions compete for the activation of the US actuator, but their comparison is impossible

Using the universal model of data meaning for all types of sensors modalities and knowledge

4

5 of 21

Cognitive Perception System functions�� � Computing the sense of data from different sensors using the universal model of meaning representation �� Representing the current state of environment by set of concepts that map the meaning of data from sensors by different abstraction levels � Representing the temporal changing of environment states (prehistory and plan action) by the sequence of concepts using cognitive model of data aging in the short-term memory��

5

6 of 21

������� FAI architecture� blueprint

6

7 of 21

Feeling Artificial Intelligence architecture blueprint. Our publications��1. А. Kargin, T. Petrenko, Feeling Artificial Intelligence for AI-Enabled Autonomous Systems, in: Conference Proceedings of 2022 IEEE Global Conference on Artificial Intelligence and Internet of Things (GCAIoT), Alamein New City, Egypt, 18-21 December 2022, pp.88-93.��2. A. Kargin, T. Petrenko, Knowledge Distillation for Autonomous Intelligent Unmanned System, in: Witold Pedrycz, Shyi-Ming Chen, Advancements in Knowledge Distillation: Towards New Horizons of Intelligent Systems, Studies in Computational Intelligence, vol. 1100. Springer International Publishing, 2023, pp. 193-231.��3. A. Kargin and T. Petrenko, “Method of Using Data from Intelligent Machine Short-Term Memory in Fuzzy Logic System,” in Conf. Proc. of 2021 IEEE 7th World Forum on Internet of Things (WF-IoT), New Orleans, Louisiana, USA, Jun. 2021, pp. 842-847.��4. A. Kargin, T. Petrenko, Spatio-Temporal Data Interpretation Based on Perceptional Model, in: Advances in Spatio-Temporal Segmentation of Visual Data, Studies in Computational Intelligence, V. Mashtalir, I. Ruban, V. Levashenko (Eds.), vol. 876, Springer, Cham, 2020, pp. 101-159.

7

8 of 21

Mapping data from sensors into its sense presented by meaning of concepts���������� Cognitive Perception System� � Cognitive Perception System

8

9 of 21

Foundation of Cognitive Perception�is� Knowledge Granule Model as semiotic triangle���Sign model represented by natural language words��External Meaning is the definition given like in explanatory dictionary by using other words or phrase��Internal Meaning or sense is numerical estimate of how sensations corresponds to the External Meaning ��

9

10 of 21

Graphic illustration of Knowledge Granule Model

10

11 of 21

 

11

12 of 21

 

12

13 of 21

 

13

14 of 21

The foundation of perception system of living beings� is� short term memory mechanism��Efficiency of decision making depends on the memory depth:�using a long history of environment changing, as a situation only at the current time is associated with a great risks ��The nature overcomes this problem by blurring the footprint of environment changes sequences:��the deeper event into the memory, the less confidence in it

14

15 of 21

 

15

16 of 21

Graphical illustration of blurring of the footprint of environment changes over time �a) At the current time, data from sensors have been� obtained to compute the pc of granules 1,2,3,4� Dark shadow corresponds high confidence pc = ±1� White shadow corresponds full uncertainty pc = 0���b) At the next after a) time, data from sensors have been � obtained to compute the pc of granules 5,6,7,8.� The pc of granules 1,2,3,4 is decreased due to aging.� Their shadow is little bit lighter.��c) At the next after b) time, data from sensors have been� obtained to compute pc 9,10,11,12,13,14 of granules.� The pc of granules 1,2,3,4 is decreased more then� granules 5,6,7,8

16

17 of 21

 

17

18 of 21

��������������� ����

18

At time t=2 data from direction have been obtained.

At times t=12 and 24 data from direction have been obtained

In real time scanning of the robot environment is performed by step-by-step positioning from right to left of the rotary platform on which the ultrasonic sensor is installed. An object is approaching to the right of the intersection. Presumed certainty of presence object in sectors , , and shown in the chart corresponding colors

Experiments with appliance of cognitive perception model

19 of 21

�����vR,vL =1.0 vR,vL =0.25 vR,vL =0.25���������vR,vL =0.01vR,vL =0.25����

19

 

20 of 21

Conclusion��1. Feeling Artificial Intelligence, as a blueprint of a new generation AI, is intended to ensure the autonomy of the operation of the unmanned system.��2. The FAI cognitive perception model implements such cognitive function as mapping of sensory information into short-term memory and processing by generalizing and abstracting keeping the data sense. ��3. The cognitive processes of aging of information stored in short-term memory are embodied in the FAI cognitive perception model as External and Internal Meaning of Knowledge Granule models.�� 4. Conducted computer experiments with the cognitive perception model confirmed the functionality of the data aging mechanism and its impact on the confidence of decision-making in AIUS.

20

21 of 21

Thank you

Questions ?

kargin@kart.edu.ua

21