BDIx (initially introduced in \cite{Ioannou2020}) is a BDI agent that is extended to utilise in Beliefs any other AI/ML techniques (e.g., Fuzzy Logic, Deep Learning Neural Networks, etc.) that gives, among others, to the agent a better understanding of the surrounding environment and the ability to prioritise the order the Desires will be executed. More specifically, as some Desires must conclude before the execution of others (i.e., because the output of one Desire can be an input to another), we allow the Desires to be assigned with priority values, ranging from 0 (lowest) to 100 (highest).
In our DAI framework which utilises BDIx agents, this priority value is estimated by using fuzzy logic (as shown in Figure \ref{fig:flowchart}) considering in its "IF-THEN" rules the current Beliefs, the values measured by the sensors of the D2D Device, any raised events (e.g., see Table \ref{Events}) and cases where the pre-specified threshold values (e.g., Data Rate Drop less than 60\%, Signal Quality Drop less than 30\%) are exceeded. Based on the assigned priority value, Desires become Intentions which are adopted for active pursuit by the agent (referred as a Goal). When the Intention is accomplished, the priority value of the associated Desire is set to zero. In addition, a Desire that will become an Intention can have multiple plans associated with it and the Desire can select an appropriate plan based on a utility function. For simplicity, but without loss of generality, in our DAI Framework we consider each Desire, and indirectly each Intention, to be associated with only one plan. It is also important to highlight here that the Beliefs and the Desires of the BDIx agent comprising the DAI framework have been extracted from the D2D Requirements/Challenges that should be realised in order to implement 5G D2D communication.
1. Agent-based Distributed
AI framework
1
Iacovos Ioannou, Christophoros Christophorou, Vasos
Vassiliou, Andreas Pitsillides**
Collaborators
NIT Goa, Pravati Swain (in Mobile Networks)
SSN institutions,Chennai, India, Prabaagarane. N
UJ, Johannesburg, Suvendi Rimer
JAIST, Kanazawa, Japan Saher Javaid (in Power systems)
**Dept. of Computer Science, University of Cyprus & Department of Electrical & Electronic
Engineering Science, University of Johannesburg (Visiting Professor)
CEII 2022. December 2022
2. The existing mobile network challenges:
i) huge bandwidth demands
ii) infrastructure and technologies cannot
fulfill data rate and connections needs
iii) new technologies and service demands
complicate 5G/6G management (e.g.
Softwarisation Augmented Reality /Virtual Reality
/Autonomous Vehicles)
2
Need to meet 5G requirements:
high data rates
low latency
low energy consumption
high scalability
improved connectivity and
reliability
improved security
And the more demanding 6G
requirements:
higher data rates and lower
latency
mass connectivity
flexibility in diverse services,
flexibility in managing and
controlling diverse networking
equipment, including softwarised
network elements and even flat
infrastructure (i.e. cell free, no BSs)
Problem Statement and Motivation
in 5G/6G
propose an agent-based Distributed AI Framework capable to tackle
these diverse needs and requirements
3. i) a fully decentralized control with virtual
resources
ii) a distributed control with independent
and autonomous systems (e.g. no dependence
on Base Stations)
iii) to be adaptable and flexible in any
situation (support dynamic environments)
iv) AI/ML support everywhere (from centralized
computing facilities to every terminal in the network)
3
NEW 5G/6G Requirements demand
4. Objective
o develop an agent-based Distributed AI (DAI)
framework that acts as a glue with any AI/ML
techniques utilized by the agents, enabled
through intercommunication among them to:
o satisfy complex problem requirements, e.g. in 5G,
6G, power systems, …
o support Self-Organization
o be Adaptable, Autonomous, Dynamic and
Flexible
Ioannou, V. Vassiliou, C. Christophorou and A. Pitsillides, "Distributed Artificial Intelligence Solution for D2D Communication in 5G
Networks," in IEEE Systems Journal, vol. 14, no. 3, pp. 4232-4241, Sept. 2020,.
I. Ioannou, PhD Thesis, A novel Distributed Artificial Intelligence framework with Machine Learning for 5G/6G communication, Sept
2021, Computer Science, University of Cyprus
4
BDI agents-based DAI are a promising technology
worth investigating to realise above
5. BDI agents-based DAI
• BDI agents-based DAI as a concept is based on
– intelligent agents that manage
their knowledge, abilities, capabilities
and intents/plans
to perform actions
with objective to solve a problem
by collaboration or as individual entities
5
Amato, A., & Venticinque, S. (2014). A distributed agent-based decision support for cloud brokering.
Scalable Computing, 15(1), 65–78. https://doi.org/10.12694/scpe.v15i1.966
• DAI scheme can support perfectly parallel workload
• This is often the case where there is little or no dependency, or
need for communication between those parallel tasks/nodes
6. Why BDI (Belief-Desire-Intention) agents?
because of their unique features
– Beliefs: correspond to the informational state of the agent
• can include inference rules + allows advance chaining to guide to new beliefs
– Desires: correspond to the motivational state of the agent
• characterize objectives or situations that agent would like to fulfil
– Intentions (what): correspond to the deliberative state of the agent
• This is what agent chose to perform by executing a plan
6
BDI agents-based DAI
These unique attributes contribute to the achievement
of desirable DAI Framework performance
Note: Intentions are desires to which the agent has commitment, and a goal is a
desire (converted to intention) that has been adopted for active pursuit by the agent
7. Extended BDI agents (BDIx agents)
BDIx agent
• can utilize in Beliefs any other AI/ML for better
understanding of surrounding environment to agent
• can prioritise the order of execution of the Desires
– Desires may depend on the completion of others to
start/conclude, thus priority values are assigned to control
their execution
– Plan Library must also be used for controlling the execution
of Desires converted to Intentions, and thus restrict agent
deliberation (note: Intentions can change at runtime and anytime)
7
I. Ioannou, PhD Thesis, A novel Distributed Artificial Intelligence framework with Machine
Learning for 5G/6G communication, Sept 2021, Computer Science, University of Cyprus
8. Summary of BDI and BDIx agents
differences
8
BDI Agent BDIx Agent
Utilises other AI/ML approaches at Beliefs N Y
Uses Fuzzy logic with priorities values on Beliefs N Y
Filters Sensor Values and Raised Events N Y
Provides REST API to Telcom Operators N Y
Has LEGO Based Components N Y
Provides Concurrent Execution of Multiple Intentions N Y
Provides ACID mechanism for Beliefs N Y
Has an Architecture for the implementation
Simpler
Architecture Y
Has a Flowchart of execution that support the above
Simpler
Flowchart Y
Enforces through the BDIx Interpreter the whole
implementation of the DAI Framework
No Supported
yet Y
Provides additional Features based on the 5G/6G requirements
Specific
Features Y
Adapts the Characteristics to be aligned with the requirements Y
9. A BDIx-based DAI Framework
9
• can achieve specific task/requirement through distributed
agent communication
• uses targeted modules within the BDIx agents-based DAI
framework
• modules can be substituted or added as needed (extra AI/ML
models)
• offers intercommunication and collaborative decisions
• by using well-defined messages of BDIx agents in the
framework
• can use many predefined well-structured languages for
BDI agents communication (e.g. FIPA ACL*)
• including “propose”, “notify”, and “inform” types of messages
*http://www.fipa.org/repository/aclspecs.html
10. constituent components, include:
– i) Beliefs
– ii) Desires
– iii) Plans, which are associated to the Desires
– iv) Threshold values
– v) Events
– vi) Plan library: handles priority of Desires to
become Intentions
• can be added or removed at run time (as long as they are not used by
the BDIx agent during the execution of Intentions)
10
BDIx-based DAI Framework
11. Basic architecture of the implemented
BDIx Agent (1/5)
13
I. Ioannou, PhD Thesis, A novel Distributed Artificial Intelligence framework with Machine Learning for 5G/6G communication, Sept
2021, Computer Science, University of Cyprus.
I. Ioannou, C. Christophorou, V. Vassiliou, A. Pitsillides, A novel DAI framework with ML for D2D Communication in 5G/6G Networks,
Computer Networks Journal (COMNET), 211 (2022).
Main modules: beliefs, desires, intentions
lead to goals and actions
12. Basic architecture of the implemented
BDIx Agent (1/5)
14
Beliefs – understanding of the environment variables
that represent the environment state
• e.g. the frequency bands, the neighbouring agents, channel quality
13. Basic architecture of the implemented
BDIx Agent (2/5)
15
Desires - : are what the agent would like to do e.g. keep
the signal quality at acceptable levels
• A desire is triggered by (filtered) events
• Desires can be prioritised to intentions
14. Basic architecture of the implemented
BDIx Agent (3/5)
16
Intentions are current desires that our
agent has committed to pursue
Many intentions may exist,
e.g. security, maximise quality, …
15. Basic architecture of the implemented
BDIx Agent (4/5)
17
BDIx agents prioritise the order of execution
– e.g. use fuzzy logic to affect the priority values of the queue
– a plan library will set the specific priority values of the desires
16. Basic architecture of the implemented
BDIx Agent (5/5)
18
How do we select which
intention will run?
• Through a queue which intentions are
slotted to run, based on their priority.
• These priorities set the order of execution
of desires
17. Extended architecture of the
implemented BDIx Agent
19
Inter-agent communication and handling unexpected events
18. The complexity of BDIx agent
implementation
• complexity only depended on the
– implementation of the Fuzzy Logic controller
– complexity of the plan that is currently executed
because all the other components are just Queues and
Lists that are part of an Object (called BDIx agent)
• based on infrequent call of Fuzzy Logic Engine
expect computational load to be rather low
– especially if Fuzzy Logic is implemented as a table
with parameters
20
https://www.sciencedirect.com/science/article/abs/pii/S0165011497004090
20. BDIx-based DAI framework
22
i) fast control with less messaging exchange (hence a
reduced signaling overhead with fast decision making)
ii) support of self-healing mechanisms and to
collaboratively act as a self-organising network
iii) can capitalise on existing mechanisms/
implementations (e.g., ANNs, optimized functions)
offers:
21. D2D as an example of utilization of the
DAI Framework
To better illustrate the concepts of the DAI
framework a Device-Device (D2D) setup in
5G is considered:
In this setup, each D2D device, aims to tackle
the D2D challenges by focusing on the local
environment of D2D communication
23
I. Ioannou, C. Christophorou, V. Vassiliou, A. Pitsillides, A Distributed AI/ML Framework for D2D
Transmission Mode Selection in 5G and Beyond, Computer Networks Journal (COMNET), 210 (2022)
22. D2D as an example of realisation of the DAI
Framework
• Operates in licensed (inband
D2D) and unlicensed (outband
D2D) spectrum
• Transparent to cellular
network
• allows proximate
devices to bypass BS
and establish direct
links between them
(share their connection, act as
relay stations, or directly
communicate and exchange
information)
I. Ioannou, V. Vassiliou, C. Christophorou and A. Pitsillides, "Distributed Artificial Intelligence Solution for D2D
Communication in 5G Networks," in IEEE Systems Journal, vol. 14, no. 3, pp. 4232-4241, Sept. 2020,
D2D communication to play a major role in the realization of 5G/6G
24
23. Challenges in D2D
• Many challenges, including:
– Device discovery
– Mode selection
– Interference management
– Power control
– Security
– Radio resource allocation
– Cell densification and offloading
– QoS / Path Selection
– mmWave
– Handover of D2D device
25
I. F. Akyildiz, S. Nie, S. C. Lin, and M. Chandrasekaran, “5G roadmap: 10
key enabling technologies,” Comput. Networks, vol. 106, pp. 17–48, 2016
24. Implementation Specifics for Meeting
the D2D Requirements
• D2D challenges were implemented with
Plans* associated with the related
Desires
• In addition, some D2D challenges
–handled when specific network Events
are raised, and/or
–some Beliefs change due to Sensor
readings or raised Events
26
*The PhD thesis outlines a plan for each of the D2D challenges
I. Ioannou, PhD Thesis, A novel Distributed Artificial Intelligence framework with Machine Learning for
5G/6G communication, Sept 2021, Computer Science, University of Cyprus
25. recall, BDIx agent characterised by:
–components
• Beliefs
• Desires
• Intentions and Goals
–and its behaviour
• Perception
• Plan
27
Implementation Specifics for Meeting
the D2D Requirements
26. 28
To handle D2D communications with BDI Agents
BDIx agent reasoning should be created.
Illustrative (limited) sets of BDIs:
– Beliefs:
• Channel, transportation mode (Relay/Cluster/Relay hop),
• technology used (5G/Wi-Fi/Bluetooth/mmWaves),
• location, neighbor agents, maximum radius, next hop (D2D relay or BS
channel)
– Desires:
• Find best Transmission Mode that achieves the best achievable
signal quality and data rate
• Data Rate is acceptable, Signal quality is acceptable
– Intentions:
• when Desire “Find Transmission Mode with best achievable signal
quality and data rate” has priority 100% it is converted to
Intention and it is moved to Goals
• In the Goals the Intention will run as in Plan
Distributed AI Solution (DAIS) for D2D
communication in 5G networks
I. Ioannou, V. Vassiliou, C. Christophorou and A. Pitsillides, "Distributed Artificial Intelligence Solution for
D2D Communication in 5G Networks," in IEEE Systems Journal, vol. 14, no. 3, pp. 4232-4241, Sept. 2020,
27. DAIS Plan For Transmission Mode Selection: new device entering
31
I. Ioannou, V. Vassiliou, C. Christophorou and A. Pitsillides, "Distributed Artificial Intelligence Solution for
D2D Communication in 5G Networks," in IEEE Systems Journal, vol. 14, no. 3, pp. 4232-4241, Sept. 2020
An example plan in a BDI agent
Step 1. Desire "Device Discovery" becomes an
Intention
Step 2. Once fulfilled, priority value of related Desire set
to 0% while rest of Desires increased
Step 3. Desire "Data Rate is acceptable" becomes an
Intention. Related Desire is associated with "DAIS" Plan
which goes through the following steps:
(a) Compute WDR of proximity devices and
select highest
(b) Set Transmission mode as "D2D Client" .
Initiate connection to D2D relay with
highest WDR.
(c) Request connection to D2D relay
(d) D2D relay responds to the request
(e) D2D Client connects to the D2D relay
Step 4. Once DAIS plan is finalized and Intention is
achieved, priority value of related Desire is set to 0%.
Then another Desire selected, if any, based on priority
values set by Fuzzy Logic rules, to become an Intention
28. Evaluation setup and scenarios
• simulation of 10 ≤ N ≤ 1000 D2D Devices
• placed in cell range of 1km radius from BS
using Poisson Point Process distribution
• simulation environment implemented in Java
using specific libraries from Matlab 2020a
"5G/LTE Toolbox” in conjunction with JADE
library
• comparatively evaluated against several
other approaches
32
I. Ioannou, V. Vassiliou, C. Christophorou and A. Pitsillides, "Distributed Artificial Intelligence Solution for D2D
Communication in 5G Networks," in IEEE Systems Journal, vol. 14, no. 3, pp. 4232-4241, Sept. 2020
29. Indicative gains by using BDIx agents
33
I. Ioannou, V. Vassiliou, C. Christophorou and A. Pitsillides, "Distributed Artificial Intelligence Solution for D2D
Communication in 5G Networks," in IEEE Systems Journal, vol. 14, no. 3, pp. 4232-4241, Sept. 2020
Simulation setup link
30. Comparative performance evaluation
34
I. Ioannou, V. Vassiliou, C. Christophorou and A. Pitsillides, "Distributed Artificial Intelligence Solution for
D2D Communication in 5G Networks," in IEEE Systems Journal, vol. 14, no. 3, pp. 4232-4241, Sept. 2020
31. DAI Framework in a dynamic environment
35
In the dynamic environment, every mobile has direction and speed.
We consider the speed of UE in the dynamic DAIS plan before selecting
transmission mode.
If speed > threshold it cannot be selected as a D2D-Relay
32. DAI Framework in a dynamic environment
36
In the dynamic environment, DAIS achieves the best results, as compared to
other approaches
33. Illustrative example conclusions
• introduced a novel agent-based DAI framework
to address the demanding management and
control challenges in 5G/6G communication
• Introduced the BDIx agent architecture and its
constituent elements
• from the D2D implementation example, shown
that the framework:
– executes control of the communication
– achieves fast decision in control
– is dynamic and flexible
– achieves a comparatively better solution (in terms of SE and PC)
– reduces messaging exchanges
37
34. Conclusion
• introduced a novel agent-based DAI framework
to address the demanding management and
control challenges in 5G/6G communication
• Introduced the BDIx agent architecture and its
constituent elements
• from the D2D implementation example, shown
that the framework:
– executes control of the communication
– achieves fast decision in control
– is dynamic and flexible
– achieves a comparatively better solution (in terms of SE and PC)
– reduces messaging exchanges
38
35. Future Work
• extensive evaluation using both simulation and a
(small-scale) test-bed
• other Plans and Intentions for tackling 5G/6G challenges
• a game theoretic perspective of the BDIx
agents can also be investigated to form a multi-agent
system in a non-cooperation environment
• framework can be enriched with new
technologies, e.g.
– D2D caching, as well as software-driven Functional
Metasurfaces and Blockchain technology
39
38. BDIx-based DAI Framework Main
Features
• Modularity: An authorized user can add or remove Desires at run time and change
the relations between Beliefs, Plan Library and Desires easily.
• Multitasking Execution: Multiple problems can be solved concurrently by the BDIx
agent with the parallel execution of multiple Intentions.
• Collaborative Environment: Through communication with the use of e.g., FIPA ACL
and LTE ProSe, the agents can coordinate and form a collaborative environment
through which they can negotiate the acceptance of a proposal by other agents and
commit to do their proposed task by considering their Beliefs and Desires.
• Logging of User Actions: The BDIx agents can Log user action in order to improve
the QoE.
• Autonomicity: The BDIx agent can act independently in order to solve a problem
and decides for the control of communication without any dependency on
information other than the local information provided by Device Discovery
(Proximity Services).
• Dynamicity: The BDIx agent supports reinforcement learning with the use of sensors
and metrics that measure the environment and updates the Beliefs according to the
representation of the environment in order to react to changing conditions of
operation.
42
39. BDIx-based DAI Framework Main
Features (Cont..)
• Flexibility: is has the ability to adapt to possible, future changes in its requirements
and react fast in a change of a situation with the use of APIs.
• DAI: It is a Distributed Artificial Intelligence (DAI) Control with the use of BDIx
agents.
• Security: Each BDIx agent can utilise well known security techniques (e.g., RSA
encryption, SSL protocol, digital signatures ) assigned in each device as tools to
increase security.
• Environmental Representation: The BDix agents can achieve an accurate
representation of the surrounding environment in the Beliefs with the use of
sensors, variables, simple data structures and with the utilization of high
complicated data structures (i.e., Neural Networks).
• Light Execution: The BDIx agent uses reduced CPU and memory resources for
executing tasks in order to run efficiently on Smartphones and Internet of Things
(IoT) hardware.
• Deliberation: The BDI agents can have an increasing freedom for selecting Desires to
become Intentions. However, in our framework the freedom is slightly restricted by
the Fuzzy Logic rules of the Plan Library of the agent.
43
40. Architectural Characteristics of the BDI
Agents
Architectural Characteristics of the BDI Agents:
– Persistence coefficients: higher persistence to continue current
actions independently and with lower persistence to be adaptable
and reactive but with inconsistent and computational costly
behaviors.
– Priority values: is the characteristic of the agent to determine the
correct intention to be used from a corresponding Desire in case of
a Believe change or the raise of an event.
– Flexibility: is the ability of the agent to easily define and adapt its
Beliefs, Desires and Intentions in real time.
– Responsiveness: is the agent's behavior and responsiveness to
events raised, sensor measured values, and changes in its Beliefs.
– Reactivity: a reactive agent can define a cognitive model and
through this model specify its target challenges along with the
plans that will achieve their implementation.
44
these characteristics are adjusted or extended to achieve
the 5G/6G requirements.
41. Relevant publications
• I. Ioannou, C. Christophorou, V. Vassiliou, M. Lestas, A. Pitsillides,
Dynamic D2D Communication in 5G/6G using a Distributed AI
Framework, IEEE Access, Print ISSN: 2169-3536, Online ISSN: 2169-
3536, DOI: 10.1109/ACCESS.2022.3182388, Vol. 10, June 2022, pp.
62772-627991.
• I. Ioannou, C. Christophorou, V. Vassiliou, A. Pitsillides, A novel DAI
framework with ML for D2D Communication in 5G/6G Networks,
Computer Networks Journal (COMNET), 211 (2022) 10898,
https://doi.org/10.1016/j.comnet.2022.108987, July 2022.
• I. Ioannou, C. Christophorou, V. Vassiliou, A. Pitsillides, A Distributed
AI/ML Framework for D2D Transmission Mode Selection in 5G and
Beyond, Computer Networks Journal (COMNET),
https://doi.org/10.1016/j.comnet.2022.108964, Computer Networks
210 (2022) 108964, June 2022.
• I. Ioannou, V. Vassiliou, C. Christophorou, A. Pitsillides, Distributed
Artificial Intelligence Solution for D2D Communication in 5G networks,
ΙΕΕΕ Systems Journal, Digital Object Identifier:
10.1109/JSYST.2020.2979044, Print ISSN: 1932-8184, Online ISSN:
1937-9234, on-line: April 2020, Print: September 2020, Volume: 14,
Issue: 3, pp. 4232-4241.
Editor's Notes
Our objective is to Develop a Distributed AI with ML framework that will act as a glue with other Intelligent Approaches/ Machine Learning (AI/ML) utilized by Mobile Devices and enable them to intercommunicate among them. The target is to achieve 5G and beyond (6G) requirements and to support Self-Orginizing Network and be Autonomous, Dynamic and Flexible
DAI as a concept based on intelligent agents that manage their knowledge, abilities, capabilities and intends/plans in order to perform actions with the objective to solve problem(s) by collaboration or as individual entity for problem solving
DAI scheme supports perfectly parallel workload. This is often the case where there is little or no dependency, or need for communication between those parallel tasks/nodes
Why BDI (Belief Desire Intention) agents?
Because of their unique features. BDI stands for:
Beliefs: Beliefs correspond to informational state of agent
Desires: Desires correspond to motivational state of agent
A goal is a desire (converted to intention) that has been adopted for active pursuit by the agent
Intentions (what): Intentions correspond to the deliberative state of the agent
These features make them to achieve the DAI benefits that we will see next!
A brief comparison of BDI agents found in the open literature and BDIx agents implemented in the thesi are presented in the table. We will examine BDIx agents in the following slides.
BDIx-based DAI Framework acts as the glue platform in employing optimised intelligent approaches, relying only on local knowledge. BDIx agents in the framework support intercommunication and collaborative decisions and use well structured language.
Lego style Components of the DAI Framework that can be changed at runtime by operator.
The purpose of the Plan Library, with the use of maximum executions of intention value and priority values is to Restrict the deliberation with the use of GOALs Queue (that supports concurrent execution of Intentions) and Intention Queue (Desires that are converted to Intentions with 100% priority value).
Additionally, we have use ACID (atomicity, consistency, isolation, durability) mechanisms targeting the correct update/delete/read of the Beliefs values though transactions locking mechanisms.
The Plan Library Pre-specifies and restricts the order of execution of Desires/Intentions with the use of Fuzzy Logic according to sensor values changes or raised of events.
At the right-hand site, we have the Execution flowchart of an Intention.
The basic DAI Framework architecture
Main modules: beliefs, desires, intentions lead to goals and actions,
supporting functionalities complete the BDIx agent
Inter-agent communication
Handling unexpected events
What is Device to Device communication (D2D)?
It is a type of communication that: i) Operates both in the licensed (inband D2D) and unlicensed (outband D2D) spectrum; ii)Transparent to the cellular network as it allows proximate devices (UEs) to bypass the Base Station and establish direct links between them (Share their connection and act as relay stations, Directly communicate and exchange information)
In this slide we provide an illustrative set of Beliefs, Desires and Intentions for the D2D DAIS Plan to run.
Centralised and distributed control
With DAI control we reduce workload at BS and make our network devices intelligent by shifting the control to the devices.
Through Transmission Mode Selection we can improve the SE and PC by using clusters (D2D Relay) and backhauling links (D2D Multi Hop Relays).
The setup of the evaluation and scenarios
What we gain with the use of BDIx Agents based DAI Framework?
We gain:
High Data Rate (as shown in the Figure 1 & 2)
Less Power Consumption therefore we reduced the energy needed for the communication drastically (as shown in the figure 3)
Reduced the computation Time and therefore we can support a mobile network that is dynamic (as shown in the figure 4)
Just show
Many aspects of DAI Framework and its utility can be tackled, e.g.
Many aspects of DAI Framework and its utility can be tackled, e.g.
Thank you!
Be Brief
This slide provides the important competitive features that BDIx-based DAI Framework supports (green color from existing literature, blue are introduced by us). We utilise existing literature findings, extended where applicable or we create our own implementation and contribute (e.g., BDIx agents, architecture, flowchart of intentions execution, REST APIs to operators).
Just show
Architectural Characteristics of the BDI Agents:
Persistence coefficients: higher persistence to continue current actions independently and with lower persistence to be adaptable and reactive but with inconsistent and computational costly behaviors.
Priority values: is the characteristic of the agent to determine the correct intention to be used from a corresponding Desire in case of a Believe change or the raise of an event.
Flexibility: is the ability of the agent to easily define and adapt its Beliefs, Desires and Intentions in real time.
Responsiveness: is the agent's behavior and responsiveness to events raised, sensor measured values, and changes in its Beliefs.
Reactivity: a reactive agent can define a cognitive model and through this model specify its target challenges along with the plans that will achieve their implementation.
Architectural Characteristics of the BDI Agents:
Persistence coefficients: higher persistence to continue current actions independently and with lower persistence to be adaptable and reactive but with inconsistent and computational costly behaviors.
Priority values: is the characteristic of the agent to determine the correct intention to be used from a corresponding Desire in case of a Believe change or the raise of an event.
Flexibility: is the ability of the agent to easily define and adapt its Beliefs, Desires and Intentions in real time.
Responsiveness: is the agent's behavior and responsiveness to events raised, sensor measured values, and changes in its Beliefs.
Reactivity: a reactive agent can define a cognitive model and through this model specify its target challenges along with the plans that will achieve their implementation.
The purpose of the Plan Library, with the use of maximum executions of intention value and priority values is to Restrict the deliberation with the use of GOALs Queue (that supports concurrent execution of Intentions) and Intention Queue (Desires that are converted to Intentions with 100% priority value).
Additionally, we have use ACID (atomicity, consistency, isolation, durability) mechanisms targeting the correct update/delete/read of the Beliefs values though transactions locking mechanisms.