LLamasoft Supply Chain Blog

← Back to Blog
  • Trucks and Cars Driving on Motorway

World Record Tied: Transportation Optimization in Supply Chain Guru X

By LLamasoft Applied Research  September 27, 2017

Earlier this summer LLamasoft launched Supply Chain Guru X, a completely new supply chain design software application. We’re proud of this new generation of our flagship product, combining the best of the best supply chain optimization technologies, with exhaustive rounds of user-feedback to offer the best solution to meet needs and objectives. As part of this latest version, we’ve also made significant improvements to our transportation optimization solving technologies. Here are just a few of the updates we’ve made to offer you the most accurate and effective outputs.

Performance Improvements

Supply Chain Guru X is our flagship transportation optimization solver and sets new benchmarks in the transportation routing world. In a recent accomplishment, our solver managed to tie the best-known solution on a large scale model used to benchmark academic and private solvers in a world record competition that encompassed a collection of 60 models compiled by two leading German research professors. Not only did our new solver capabilities solve in ten minutes, faster than the other datasets, it also offers a more broad and scalable application. Over the past year we re-built the solver from the ground up by code refactors, algorithmic enhancements and consistent definitions. LLamasoft customers upgrading to the latest version of Supply Chain Guru may feel like they are upgrading from an old pick-up to the sleekest, fastest and most reliable sportscar on the market today.

If that doesn’t convince you that the new solver is awesome, the following metrics should. The performance improvements were computed by running 200 models that cover a wide variety of use cases in terms of functionality and problem size. The run times are only computed for models that take more than 30 seconds to solve to eliminate noise. The overall performance improvements on models from the previous to the current version of Supply Chain Guru include:

  • Standard Optimization:
    • Total Cost lowered by 10.66% – 9.93% for models without Fleet Optimization and 11.16% for models with Fleet Optimization
    • 15% faster –14.59% for models without Fleet Optimization and 15.15% for models with Fleet Optimization
  • Interleaved Optimization (in which deliveries and pick-ups are interspersed):
    • Total Cost lowered by 24.73% – 26.4% for models without Fleet Optimization and 23.98% for models with Fleet Optimization
    • 24.82% faster – A whopping 56.7% for models without Fleet Optimization and 4.89% for models with Fleet Optimization
  • Hub Optimization:
    • Total Cost lowered by 6.64%
    • Expect similar run times

Major Algorithmic Enhancements

  1. Cost-based asset selection in Standard Optimization

In the previous version of Supply Chain Guru, the Standard Optimization solver selected assets either based on user defined priority order or a set of deterministic rules. This occasionally meant the best option never even made its way to the table due to a self-selection oversite. In the latest version, the solver now generates routes across all asset types available to select the set of routes that minimize total transportation cost. Note: This results in a higher complexity, specifically for models with lots of assets defined, and thus longer solve times. However, the speed increase gained in other algorithmic improvements is typically significant enough to perform these additional operations in a similar or faster time (Standard models still run 15% faster in the latest version, on average).

  1. Mode Selection (Direct Shipping Cost trade-off) in Standard Optimization

We know our customers wanted mode selection and we’re happy to offer this enhancement in the newest version! The solver is now smart enough to tell you whether you should route a shipment or ship it direct by considering this trade-off as part of the decision variables.

  1. Warm-Start Baseline

You already knew transportation optimization is built on heuristics guided by optimization. In those rare occasions where the heuristics do not identify a better solution, Warm Start comes to the rescue. It lets you guide the solver by initializing some routes you think are good options. The solver will then try to improve on them if possible. (If you don’t see an improvement, you’ve built a solution that is darn good. If you built it from scratch you might want to send us your resume, we’re always looking for more talented folks!)


Modeling Enhancements

  1. More flexibility when modeling Rates

Users now can use rates as constraints by adding a custom parameter ‘UseRatesAsConstraint’ and setting it to TRUE in Config_TO1. The solver will only build routes that it can find a populated rate record for. Additionally, for Origin based rates i.e. “Origin Dependent”, “Origin – Last Destination” or “Origin-Furthest Destination”, ‘Origin’ refers to first pickup location in the route (even if this is different from asset home site). Users can direct the solver to treat asset home site as ‘Origin’ by using a custom parameter ‘UseAssetSiteAsOrigin’ set to TRUE in Config_TO1.

  1. More flexibility when modeling Business HoursAn asset reaches a customer location at 4:50 PM. The customer location closes at 5 PM as per business hours table. The unload time is 20 minutes. Based on who we ask, some users want this to be feasible while some don’t. By default, we now require the service to finish before location closes. However, we give users the choice to continue servicing a closed site if the service time started while the site was open by using a custom parameter ‘ServiceTimeEndInTimeWindow’ that can be entered in Config_TO1
    1. Consistent Definitions Across All Problem Types

    If you never experienced inconsistent behavior for some fields when using different problem types – great! If you did, you won’t anymore. Key fixes include consistency while calculating stops in a route, service times and rates2

Simplified Run Screen Interface

We’ve streamlined the run screen in the latest version to offer you the most commonly used parameters. For the expert users that liked tweaking those parameters, you can still change them from Config_TO1.

  1. Solver Setting Slider with five settings

Want quick results for prototyping? Want the solver to spend more time to find a better solution? The solver setting slider bar is your control for solution quality vs solve time tradeoff. This trade-off is achieved by controlling heuristic search space. As the search space increases, so does the runtime and the likelihood of finding a better solution. Under the hood, this slider changes the values of multiple parameters in different parts of the code to collectively control the heuristic search-space (Remember, advanced users can still control individual parameters through Config_TO1).

By default, solver setting is ‘Balanced’ which, like the label suggests, provides a good balance between search space and solve time. Typically, when first setting up your model, you may want to use the fastest setting to quickly assess the solution. As you refine the model, you can adjust the setting to expand searches during optimization.

  1. Auto-Detect Problem Type

You don’t need to select a problem type anymore. The Auto-Detect module analyzes your model structure to determine the appropriate Transportation Optimization Problem. The detected problem type will clearly appear on the run screen log. You can also choose to override it from the Advanced tab and select a specific problem type you want.

With version X, Transportation Optimization at LLamasoft continues to show its commitment to users by adapting to their needs and providing the best solvers, features, and support. We measure our success in the success of our customers and hope you are as excited about Supply Chain Guru X as we are!

Want to learn more about transportation optimization? Check out our new ebook, “7 Tips from the Experts for Creating a Competitive Advantage with Transportation Optimization,” with insights from Ryder, Schneider Electric and more.