Behavioral Micro-Simulation 1 Jose Holguin-Veras, Ph.D., P.E. William H. Hart Professor VREF’s Center of Excellence for Sustainable Urban Freight Systems Center for Infrastructure, Transportation, and the Environment Rensselaer Polytechnic Institute
Main goals To produce a reasonable guess of freight traffic in metropolitan areas using: Freight trip generation estimates (using NCFRP 25 models) Known delivery patterns, such as tour length distributions by industry sectors (obtained from data collected by RPI from carriers and receivers) Observed traffic counts at key corridors The BMS was originally developed to assess the impacts of alternative policies to foster off-hour deliveries (7PM to 6AM) 2
Key components Freight trip generation (FTG): estimated using the NCFRP 25 models and Zip Code Business Pattern data Synthetic population of carriers (and receivers, if needed) is created Using the data collected by RPI, the sample data is used to create the population of carriers needed to make all deliveries in the metro area The origin of the deliveries are set to be the locations were warehouses and distribution centers are located Delivery tours are created: Match the tour length (number of stops) by industry sector Match the number of deliveries by ZIP code (or any other level of geography used) 3
Graphically: Freight Trip Generation 4
Graphically: Synthetic population of carriers Different industry sectors have different tour lengths NYC and NJ (Holguin-Veras et al. 2012): Average: 8.0 stops/tour; 12.6% do 1 stop/tour; 54.9% do 20 stops Synthetic population match observed traffic and FTG 5
Tour simulations Select a truck in an industry sector Number of stops is randomly assigned Select receivers at random from the group of receivers in that sector Compute optimal tour and store it Repeat until delivery tours satisfy the FTG for the entire area 6 1) Origin of a truck that carries food products to five restaurants 2) Five receivers
Example: Use of the BMS in the OHD project
BMS use in the off-hour delivery project 8 Carrier/receiver synthetic generation Randomly select industry segment o Generate/locate carrier o Generate/locate receivers to serve Receiver behavioral simulation Model receiver’s decision to accept OHD Carrier behavioral simulation Compute costs for base case and mixed operation Model carrier’s decision Repeat for another carrier-receivers set End Change incentives, reset participation counts Define range of incentives to receivers for OHD Ordinal logit model (Holguin-Veras et al 2013) Regular-hour receiver Off-hour receiver a) Base case (no OHD) b) Mixed operation Carrier depot Legend: Output: Joint Market Share (JMS) of OHD Receivers Market Share (RMS) at TAZ level
Ordered logit model with random effects This model reproduces receivers’ response to incentives 9 Incentives Interaction terms: OTI and NAICS NAICS code Interaction terms: TV and NAICS
BMS Results 10 OTI = $0 avg = 2.2% max = 6.2% min = 0.6% OTI = $2,000 avg = 2.7% max = 7.6% min = 1.2% OTI = $4,000 avg = 3.4% max = 7.6% min = 1.3% OTI = $6,000 avg = 4.3% max = 9.9% min = 1.9% OTI = $8,000 avg = 5.5% max = 11.9% min = 2.6% OTI = $10,00 avg = 7.0% max = 13.4% min = 3.5%
Example: Geographically Oriented Incentives
Geographically focused incentives: case of NYC 50% of establishments are located in Midtown Manhattan being responsible for 52% of the incoming freight trips to the city Two geographic distribution have been considered: (1) Lower and Midtown (2) Central Park and Upper Scenarios consider giving incentives to either the entire Manhattan or only to Lower and Midtown Manhattan Lower Manhattan (LM) Midtown Manhattan (MM) Central Park (CP) Upper Manhattan (UM) + +
Results of geographically focused incentives Ratio Budget/JMS provides an idea about the amount of resources required to achieve a 1% JMS The results also show the superiority of geographically focused incentives which requires between 71% and 75% less expenditures than incentives spread out all over Manhattan 13
Example: Self-Supported Freight Demand Management
Self supported freight demand management A self-supported freight demand management system (SS-FDM), is one that generates the funds required for a continuing improvement towards sustainability The incentives to be handed out to the receivers are generated by a toll surcharge to the vehicles that travel in the regular hours The analyses consider tolls to only trucks (per axle) or both; trucks and cars. Finally, different levels of toll collection efficiency were also considered 15
Results: tolls to trucks (per axle) Toll collection 100% Toll collection 75% 16 Note: The shaded cells represent non-feasible combinations of financial incentives to receivers and tolls.
Results: tolls to trucks (per axle) and cars Toll collection 100% Toll collection 75% 17 Note: in this case all combinations of financial incentives to receivers and tolls are feasible
Potential Uses
The BMS will replicate freight traffic in any metro area The BMS could be used to: Produce realistic estimates of freight VMT Analyze the impacts of alternative logistical configurations (using a Urban Consolidation Center, transfers of cargo to environmentally friendly modes like freight bicycles) Analyze the impacts of retiming of deliveries, or receiver-led consolidation programs by receivers Analyze the impacts of policies that change operational patterns, technologies, or infrastructure used by carriers Changes in work hours, limited emission zones, etc. 19
Expected outputs of the BMS Acceptance rate of technology/ operations/ infrastructure in response to policy measures Freight (large and small trucks) VMT by industry segment for the initiatives considered, including time of day for some Freight traffic by origin-destination before/after, a key input for traffic simulation models Cost impacts on carriers and receivers 20
Limitations Estimation of air pollution The BMS is not a traffic simulator, it does not account for traffic behavior in networks Potential solution: Use the BMS output as an input to traffic simulators Purchase GPS data for key metro areas and post-process it with MOVES to produce estimates, add the estimates to BMS The BMS is very good for urban freight modeling, though it does not consider intercity freight (and things like truck stop electrification, etc.) Potential solution: create modules that perform these computations, add to BMS 21
Conclusions
The BMS is an important tool to evaluate TDM policies The application to the Manhattan case study provides insight into the potential benefits, and limitations: Off-Hour Deliveries Geographic oriented incentives Self Supported Freight Demand Management Other extensions of the BMS include the analysis of incentives according to industry segments 23
Questions? 24