PACK Truck Digital Twin · EU road freight multi-agent orchestration
Scenario: Baseline EU Corridors · DEMO-SCALE assumption
T+00:00 · 0 events
Truck
Border / Customs
Hub / Terminal
Port
Transport corridor

Simulation Results · Time-Series KPIs

Charts use accumulated values from the current simulation run. Stop fixes the final run state; Initialise starts a new reproducible experiment.

Delay and border waiting time

● Average delay● Border wait

Operational throughput

● Delivered● Inspections● Risk requests

Corridor utilisation

● Network utilisation

Risk Scoring quality

● Precision● Recall● FPR● FNR

Independent HUNTER Risk Scoring Service

The service is logically separated from PACK Truck. A simulation uses at most one logical Risk Scoring service. Redundant infrastructure is represented only as High Availability support, not as additional scoring agents.

TransportScoreCategoryProvenanceDecision

Experiment Comparison

Store two completed or in-progress runs and compare them using common KPIs. Identical seeds and configurations support reproducibility.

KPIRun ARun BDifference

PACK Truck Multi-Agent Architecture

Strategic User Agent
Defines objectives, scope, legal environment, disturbances, autonomy and KPIs.
Scenario Management Agent
Initialises network, demand, populations, congestion and scheduled what-if events.
Operational Agents
Truck, Driver and Cargo agents move, rest, queue, load, unload and accumulate transport history.
Organisational Agents
Logistics Providers and Infrastructure Operators assign resources, reroute flows and manage capacity.
Legal & Regulatory Environment
Configurable jurisdictional rules are retrieved by Customs and Regulatory agents rather than hard-coded into individual agents.
Regulatory / Customs Agents
Use documentation, policy, capacity, anomaly information and Risk Scoring as decision support.
Independent HUNTER Risk Scoring Service
One logical service per simulation. Receives current and historical transport context and returns score, category, uncertainty, factors, timestamp and model version.
HUNTER Analytics
Anomaly Detection and Routing remain distinguishable components. Their outputs may feed Risk Scoring or operational decisions.
Game AI & Strategy Search
Searches difficult scenarios and policy alternatives without silently modifying the selected legal environment.
Feedback & Learning Dataset
Stores labelled outcomes with REAL, SIMULATED or DERIVED provenance. Retraining is a controlled separate process.
Deployment Abstraction
The demonstrator is standalone and deployment-neutral. The same logical services can later be containerised for local, on-premises or AWS development without making the simulation dependent on AWS-specific functionality.

Event Bus

Transport, infrastructure, regulatory and HUNTER events are time-ordered. The simulator may process them sequentially while preserving causal relationships.

Environment state
Normal operations
Risk Scoring service
Healthy · 1 logical service
Latest event
Ready for initialisation