Lessons Learned Using ESRI s Network Analyst to Optimize Snow Treatment Routes in Kentucky Ben Blandford, PhD Eric Green, PE, PhD Candidate Kentucky Transportation Center University of Kentucky
Overview Background of project Selection of NA and VRP How we set up VRP Pacman problem Multiple/parallel line segments Constraining parameters Problems with VRP Results not implementable Results not optimized? Creating route trees Results Pendleton/Grant Clark/Montgomery Successes Elimination of contract trucks Further research and implementation
Background KYTC typically expends between $40-$80 million/year $75 million in 2014 (not so much FY 2017) Routes are largely county-based, with each county having a designated number of trucks and facilities available to address the county s needs Additionally, these routes are prioritized into categories according to several factors, with Average Annual Daily Traffic (AADT) being the most influential Interstates receive the highest priority
Background The currently used snow and ice removal system has been in place for some time and functioned effectively GIS-based analytical tools could be used to further optimize the system System optimization could potentially improve mobility and safety by clearing roads more efficiently If this potential is realized, KYTC could also save considerable time and money through system optimizations
Research Objectives Utilize ArcGIS to analyze the current system of snow and ice Determine if any optimizations to the system are possible This analysis includes the following factors: Routes Priority Lanes Location of salt and brine facilities Distribution and type of trucks Time required for road clearance
Esri Network Analyst and Vehicle Routing Problem Network Analyst network-based spatial analysis tools for solving complex routing problems plan routes for an entire fleet, calculate drive-times, locate facilities and solve other network related problems Vehicle Routing Problem goal is to best service the orders and minimize the overall operating cost for the fleet of vehicles the VRP solver finds the best routes for a fleet of vehicles to service many orders the VRP solver can solve more specific problems because numerous options are available, such as matching vehicle capacities with order quantities, giving breaks to drivers, and pairing orders so they are serviced by the same route
Solving for typical winter storm Interstates and parkways should be treated during a routine winter storm with a goal of a one-hour turnaround All other Priority A routes should be treated with a goal of a two-hour turnaround Priority B routes should be treated with a goal of a four-hour turnaround Priority C routes should be treated with a goal of an eight-hour turnaround
Setting up the Vehicle Routing Problem Lanes Parallel lines created to account for multiple lanes Must account for divided highways, one-ways, ramps, etc. Delivery points Snow and ice treatment as a delivery optimization problem Midpoints created for all road segments, offset to the right of the driving direction Points represent deliveries of salt, with salt volume corresponding to length of segment Fleet Truck have salt capacities, must refill when capacity is spent Time windows created where salt deliveries must be made on schedule
Attributes of Vehicle Routing Problem Time Windows match up to route priority Salt capacity of trucks need to structure refills into the routing Location of facilities single county or multi-county? Truck type and road width Multiple lanes on routes Truck travel speed Simplifying as much as possible Input from drivers treating the routes Location of local politicians houses
Pac-Man is fun
Considerations Routing issues Desire for most efficient routes? Or desire for most realistic/efficient routes? Tricking the solver to provide more realistic routes Iterative process: altering the constraints a little can lead to very different results Time violations (H, M, L), Excess transit (H, M, L) Altering start time according to SNIC priority Grouping routes according to SNIC priority Need to reduce complexity of network as much as possible Can reduce model run times from hours to minutes or even seconds
Problems with VRP solutions Overly complex routes for fleet Truck drivers like to be familiar with their routes Truck driver on SAC committee: I would throw this map down and quit KYTC prefers simpler and more easily communicable routes Not clear that VRP routes are mathematically optimized
Route Trees Addresses the problem of needing to structure truck routes more simply Structures each route around a central high priority road Each time that road is treated, truck also treats a subset of the adjacent roads Drivers can become very familiar with their areas and routes Also makes it easier for drivers to address inevitable complexities of non-typical winter storm
State and Contract Truck Routes (Non-Interstate/Parkway) 1, A 1, A 1961 1331 2, C
State Route 1 (Tandem) 1, A 1, A 1331 1. Treat US from US to KY 57. Treat KY from US to Bath County line. Treat KY from US to KY.
State Route 1 (Tandem) 1, A 1, A 1331 2. Treat US from US to KY 57. Treat KY from US to Bath County line. Treat KY from KY to US. Treat KY 3289 from KY to Bath County line. Refill salt in truck.
State Route 1 (Tandem) 1, A 1, A 1331 3. Treat US from US to KY 57. Treat KY from US to Bath County line. Treat KY from KY to KY 57.
State Route 1 (Tandem) 1, A 1, A 1331 4. Treat US from US to KY 57. Treat KY from US to Bath County line. Treat KY from US to KY.
State Route 2 1331 1. Treat US from US to the Menifee County line. Treat KY from US to the Powell County line. Treat KY from US to KY. Refill truck with salt. 2, C
State Route 2 (single axle) 1331 2. Treat US from US to the Menifee County line. Treat KY US to the Powell County line. Treat KY from US to the Powell County line. Refill truck with salt. 2, C
State Route 2 (single axle) 1331 3. Treat US from US to the Menifee County line. Treat KY from US to the Powell County line. Treat KY 1314 from KY to End of State Maintenance. Treat KY from KY to KY. Treat KY 3363 from KY to KY. Refill truck with salt. 2, C
State Route 2 (single axle) 1331 4. Treat US from US to the Menifee County line. 2, C
State Route 3 (single axle) 1331 1. Treat KY from US to the Powell County line. Treat KY from US to the Menifee County line. Refill truck with salt.
State Route 3 (single axle) 1331 2. Treat KY from US to the Powell County line. Treat KY from KY east to KY. Treat KY southwest and back around to KY. Refill truck with salt.
State Route 3 (single axle) 1331 3. Treat KY from US to the Powell County line. Treat KY from US to the Menifee County line. Refill truck with salt.
State Route 3 (single axle) 1331 4. Treat KY from US to the Powell County line.
State Route 4 (single axle) 1331 1. Treat KY (slow lanes) from US to US. Treat US from I 64 interchange to the Bath County line. Treat KY (fast lanes) from US to US. Treat KY 1331 from US to the Bath County line. Refill truck with salt.
State Route 4 (single axle) 1331 2. Treat KY (slow lanes) from US to US. Treat US from I 64 interchange to the Bath County line. Treat KY (fast lanes) from US to US. Treat KY from US to US. Refill truck with salt.
State Route 4 (single axle) 1331 3. Treat KY (slow lanes) from US to US. Treat US from I 64 interchange to the Bath County line. Treat KY (fast lanes) from US to US. Treat KY 1331 from US to the Bath County line. Refill truck with salt.
State Route 4 (single axle) 1331 4. Treat KY (slow lanes) from US to US. Treat US from I 64 interchange to the Bath County line. Treat KY (fast lanes) from US to US.
Contract Route 1 1961 1331 Treat US from KY in Montgomery County to KY 627 in Clark County and back. Repeat and refill at either county salt barn. As time allows, treat the following: KY 1961 from US to End of State Maintenance KY 2346 from US to the airport.
Successes For two counties in District 6 (Grant and Pendleton), the model demonstrated the potential to eliminate the need for six contract trucks, which would result in a cost savings of $25,000 per truck per year. Similar results were found for two counties in District 7 (Clark and Montgomery), where route optimization demonstrated the potential to eliminate the need for three contract trucks, with a similar cost savings of $25,000 per truck per year. Within just these four counties then, there is a potential cost savings of $225,000 per year. Continued application of the model for more counties throughout Kentucky will lead to further optimization and cost savings for KYTC.