SimMobility Freight

Project Title: 

SimMobility Freight: A multi-scale agent-based simulation platform for urban freight distribution



SimMobility is the simulation platform of the Future Urban Mobility Research Group at the Singapore-MIT Alliance for Research and Technology (SMART). It integrates various mobility-sensitive behavioural models with state-of-the-art scalable simulators. SimMobility aims to enable Future Mobility research evaluations, by simulating impacts of a portfolio of technology, policy and investment options under alternative future scenarios. For example, it allows predicting the impact of intelligent transportation services on mobility demands / transportation networks. SimMobility is capable of simulating millions of agents, from pedestrians to drivers, and entities, from phones and traffic lights to GPS, moving in cars, buses and trains. The simulation time-frame can range from second-by-second to year-by-year.

SimMobility has been extended with Freight movement simulation capabilities. It simulates interactions between firms/establishments, freight vehicle drivers, and the surrounding environment in several stages of the goods distribution system. Thus, the simulation platform links economic activities, logistics choices and transportation impacts. For this, we create and make use of a synthetic population of Shippers, Carriers, Receivers, Vehicles and Shipments. The platform achieves a consistent representation of agents throughout the simulation, which allows propagation of decision-making impacts. SimMobility Freight allows exploring a new set of technology (e.g. freight parking reservation) and policy scenarios (e.g. night deliveries) as well as passenger and freight vehicle interactions. Models are compatible with innovative data collection methods.

Figure 1. SimMobility Freight Conceptual Framework


The alpha implementation of SimMobility freight consists on Long-term and Mid-term demand simulation models, illustrated in Figure 2 and described further below.


Figure 2. SimMobility Freight Alpha Implementation of Long-term and Mid-term demand simulation models.


Long-term models

  • Firm Synthesis, relies on a multi-step method to creates a population of business establishments with relevant attributes: establishments’ locations, industry type, function, employment size, occupied floor area,  fleet size (i.e. total vehicles) and constitution (i.e. vehicle types and characteristics).

  • Annual Commodities Production/Consumption (i.e., goods shipped/received) is mainly dependent on the business employment and combination of business type/supply chain function (e.g., food industry & manufacturing). Using regression models, we predict total value of annual production/consumption regardless of commodity type.

  • Commodity Types and Quantity uses local economic data ('Supply and Use' tables) to disaggregate the annual flows into commodities produced/consumed by establishments.

  • Supplier Selection matches suppliers to receivers by sequentially running two sets of discrete choice models. The first set of choice models assume the receivers to perceive the attractiveness of each zone based on: (a) the quantity of the commodity produced in it, (b) the number of potential supplier-establishments in it, and (c) the distance between the centroids of the supplier’s zone and the receiver’s zone. Coefficients to the above parameters are considered specific to commodity-type and the receiver’s position in the supply-chain. Secondly, the utilities of all suppliers in each zone are computed based on the non-allotted quantity as the simulation progress.

Shipment Size & Frequency models assume a uniform distribution of shipments throughout the year. Consist on drawing a shipment size based on a subset of records dependent on commodity type, supplier industry, and carrier type (own, hired). Use annual flow and shipment size to calculate expected frequency.




Figure 3. Illustrative geographic spread of commodities shipped and received in a Singapore application case-study.  

Mid-term models

  • Day Selection picks a subset of shipments to be simulated on a given day, based on a random draw from the probability of occurrence.

  • Carrier Selection and Carriers’ Operations Plannings consist of a sequence of heuristics pairing vehicles and shipments in a two-step process. First, suppliers who are also carriers assign as many shipments as possible to own-fleet of vehicles. Secondly, unassigned shipments are pooled with all other shipments, i.e. from suppliers who are not carriers. These shipments must be outsourced to for-hire carriers. The model iterates over the lists of shipments (first for own-account and then for the pooled remaining shipments), triggering a series of subroutines up to exhaustion of vehicle availability/capacity or shipments. These routines include, among others, the following models: (a) vehicle-selection model, (b) shipment-clustering and vehicle loading model, (c) stop-sequencing model, and (d) stop-time model.

Figure 4. Illustrative freight vehicle flows in a Singapore application case-study.

Work in progress

  • Long-term models

    • Imports/exports

    • Supplier selection and shipment size alternative formulations

  • Mid-term (Pre-day)

    • Route choice

    • Multi pickup-up delivery with time-windows

    • Supply feedback loop

    • Calibration

  • Mid/Short-term (Within day)

    • Re-routing

    • Tour re-scheduling

    • Parking choice

  • Overall
    • Consolidation models

    • Overnight parking choice models