MPAV | Jin Gao

Agent-based mobility simulation visualizing building usage, traveler types, and mode share across the airport region network.

Abstract

The goal of this multi-agent simulation is to predict a low-carbon and efficient transportation system to shift the usage and reliance on high fossil fuel consuming vehicles to multi-purpose vehicles, to shape what an optimal transportation system in the future will look like. To accomplish this, we will introduce a Multi-Purpose Autonomous Vehicle (MPAV).

Based on a modular design strategy, the MPAV is able to do tasks such as cleaning waste and delivering commercial goods. For the simulation, we simplified it to carrying people and goods(parcels and food) with two capacities. stakeholders.

We choose Kendall Square as a representative site for the future entrepreneurship innovative city, for it gathered a large number of high-tech enterprises that generate huge demand for parcel and food delivery from companies and commuters. Through simulation, we hope to find a balance between existing transportation options, MPAVs, and potentially other future modes of transportation to meet the transportation demands of the people of Kendall Square.

Method and findings

The Kendall Square MPAV study runs an agent-based model in GAMA: GIS and demand inputs feed trip and mode choices, loops update agent positions and outcomes, then results flow to dashboards and a web stack for exploration. Below, the pipeline is shown end to end alongside the MPAV versus miniMPAV capacity assumptions and the map legend used in runs.

GAMA model architecture diagram (data input → simulation → visualization) — GAMA pipeline from GIS and demand inputs through agent-based simulation to dashboard and web visualization.

MPAV and miniMPAV prototypes and capacity assumptions — MPAV ($50k, 4 people + 10 goods) vs. miniMPAV ($10k, 1 person + 3 goods) capacity assumptions.

Legend and simulation snapshot showing people + goods + MPAVs — Map legend and snapshot showing passengers and deliveries carried together on an MPAV.

Agent-based Traffic Modeling

Zoomed view of MPAV and multimodal traffic flows across the Kendall Square street network.

City-scale agent-based simulation of MPAV and multimodal traffic flows across the Kendall Square network.

Agent spawn locations in Kendall Square — Commuter entry points and food/package delivery spawn locations across Kendall Square.

Experiments

Base, MPAV, and MPAV + Policy scenario definitions — Fleet scenarios compared: baseline mobility, MPAV fleet sizes, and policy incentives.

Results

Conclusions

All MPAV based solutions save ~1000 kgCO2-eq even if the average travel time for vehicles increases by 1 min (5 min → 6 min)
All solutions involving MPAVs (except 1 MPAV) drastically reduce delivery times by ~1 hour
Optimal solution based on the above and fleet cost is 5 miniMPAVs ($10k × 5 = $50k) due to the small area of Kendall Square