Multi-Agent Pickup and Delivery in Dynamic Environments

Tech talk with Bendetta Flammini, research fellow at Politecnico di Milano.

Wednesday, 26 November | 3:00 PM CET | Online event

Autonomous multi-agent systems are becoming central to logistics, urban delivery, and warehouse automation. From fleets of delivery robots navigating busy city streets to warehouse agents coordinating item transfers, these systems must handle environments that change rapidly and unpredictably.

Yet most classical models used to coordinate such agents assume the opposite: static conditions, predictable behaviour, and homogeneous teams. These idealised assumptions limit the impact of autonomous agents in real-world scenarios where movement is uncertain, humans and robots share space, and tasks evolve continuously.

In this upcoming tech talk, Benedetta Flammini, Research Fellow at Politecnico di Milano, will present two important advances that bring Multi-Agent Pickup and Delivery (MAPD) research much closer to real operational needs.

Moving Beyond Idealised Multi-Agent Systems

Real environments are shared with unpredictable entities—pedestrians, other robots, dynamic obstacles. Benedetta introduces two behavioural models:

• Probabilistic occupancy, predicting how external agents might move.
• Markovian behaviour modelling, capturing movement patterns over time.

These models allow task-oriented agents to anticipate external dynamics and adjust their paths. The framework is paired with a new coordination algorithm that supports real-time adaptive planning and robust deadlock prevention, even in crowded settings.

MAPD with Mobile Pickups (MAPD-MP)

Many systems rely on heterogeneous agents with different roles—some act as suppliers, others as deliverers. In this extension, both agent types move through the environment under spatial and timing constraints.

To support fluid coordination, Benedetta proposes an algorithm that:

• synchronises movements between suppliers and deliverers
• optimises meeting points for item exchange
• reduces delays and increases throughput in mobile delivery workflows

Together, these contributions make multi-agent systems more capable, safer, and more aligned with operational needs in logistics, robotics, and automation.

About the Speaker

Benedetta Flammini is a research fellow at Politecnico di Milano. She obtained her PhD in Data Analytics and Decision Sciences in April 2025. Her work focuses on planning for autonomous multi-agent systems operating in changing environments, with a strong emphasis on Multi-Agent Pickup and Delivery problems.

Her research integrates environmental dynamics directly into agent decision-making processes, improving adaptability and robustness. During her PhD, she also collaborated with industry partners to analyse user behaviour data, with a focus on segmentation and behavioural insights.

Register for the Event

Join us on Wednesday, 26 November at 3:00 PM CET for an in-depth session on the future of multi-agent systems.