Please enable JavaScript.
Coggle requires JavaScript to display documents.
Distributed Artificial Intelligence, Distributed Artificial Intelligence –…
Distributed Artificial Intelligence
Distributed Artificial Intelligence – Big Picture
Definition & Need
Shift from centralised IT to decentralised, autonomous agents
Handles coordination complexity, improves scalability & robustness
Enables local autonomy and faster reaction
Key Applications
Smart Factories (Industry 4.0)
Intelligent Transport & Logistics
Autonomous Vehicles
Recommendation Systems
Digital Markets & Platform Coordination
Agent Fundamentals
Weak Notion of Agency (Wooldridge & Jennings)
Autonomy • Reactivity • Proactivity • Social Ability
Architectures
Reactive vs Deliberative Agents
Goal-Directed / Utilitarian Agents
Practical Reasoning
Deliberation → Means-Ends Reasoning (Planning)
BDI Model (Beliefs – Desires – Intentions)
Commitment Strategies & Reconsideration (Bold vs Cautious)
Communication in DAI
Speech Act Theory (Locution / Illocution / Perlocution)
FIPA-ACL & Message Structure
Interaction Protocols (Contract Net, Query, Request, custom IPs)
Communication as Computation (Theory of Mind reasoning)
Coordination & Negotiation
Criteria for Mechanism Quality
Efficiency • Individual Rationality • Stability • Simplicity • Distribution • Symmetry
Social Choice & Voting
Plurality • Approval • Cumulative • Borda • Condorcet Paradox • Strategic Voting
Social Welfare (Elitist | Egalitarian | Utilitarian | Bernoulli-Nash) + Pareto Front
Auctions & Bargaining
English vs Dutch Auctions
Multi-Issue Bargaining
Winner’s Curse
Combinatorial Auctions & Winner Determination Problem
Teamwork Model (Wooldridge)
Recognition → Team Formation → Plan Formation → Team Action
Formation Protocols (Directory, Broker, Multicast)
Emergence (Positive & Negative effects)
Decision Making & Learning
Markov Decision Processes (MDPs)
States • Actions • Transition T • Reward R • Markov Property
Reinforcement Learning
Sample → Estimate → Improve Cycle
Monte Carlo vs Temporal Difference
Q-Learning + Exploration-Exploitation (ε-greedy, Softmax)
Multi-Agent RL
Stochastic Games (Joint Actions, Agent-Specific Rewards)
Nash Equilibrium & Optimal Policies
Joint-Action Learning & Shared Q-Functions
Concurrent Q-Learning Exploration Dilemma
Infinite exploration with vanishing exploration
Trust & Reputation Mechanisms
Need for Social Mechanisms in Open MAS
Trust Definitions & Asymmetry
Trust Sources (External • Structural • Experiential)
Metrics (I⁺ / I⁻, MDT-R update, decay)
Recommendations & Transitivity
Centralised Reputation Systems (eBay example + biases)
ReGreT Model (Witness • Neighbourhood • System dimensions)
Digital Twins & Agentic AI
Digital Twin Concept (Data Loop, MAS Analogy)
Distributed DTs for Autonomous Traffic Control
Sensors • Controllers • Local Teams • Failsafe Operation
Surrogate Modelling & AutoML
Automation Checklist • When to Retrain
Embedding surrogates inside autonomous agents
Learned Futility Study
Social Influence vs Reinforcement Learning
Cost vs Effectiveness in Behaviour Adoption
Steel-Industry Decision-Support Twin
Transformation challenges & DSS requirements
Layered framework (Data • Coordinator • Model Repo • Interface)
Agentic AI Vision
From task-specific agents → compositional services → self-directed agentic systems