Go Catch ‘Em All Ann Arbor: Using Transportation Optimization for Pokémon Go
If you’ve walked down the street recently chances are good you’ve seen throngs of people playing the new augmented-reality game Pokémon Go. A location-based game, this opportunity seemed ripe for optimization. So we at the LLamasoft Applied Research Team decided to leverage the transportation optimization functionality within Supply Chain Guru, to find the best route to cover all the Poké Stops here in our backyard in the Ann Arbor area.
Whereas in a traditional transportation optimization model we’d be seeking the most optimal route to serve all locations and customers based on constraints both aligned with organizational goals and external factors, here we were modeling to find the most effective way to visit all of the Poké Stops (places where you can collect free items and potentially catch Pokémon) and Poké Gyms (places where you can train Pokémon), not yet focusing on particular constraints. Within hours, we pulled geographic data for 561 locations around the downtown Ann Arbor area. While these are supposed to cover all the Poké Stops/Gyms ) in the area, there might be a few additional locations not part of the Pokémon universe since we grabbed this data from an older game, Ingress, developed by the makers of Pokémon Go. (See  for more information).
First, as speed is of the essence in the Pokémon journey, we wanted to find the shortest path to cover all the stops. Given the co-ordinates, our transportation optimization solver automatically populated the distance matrix containing point-to-point distances for all Stops. We inflated it by 17 percent to account for actual on-road distances, based on real world applications of our software.
We built a transportation network model in Supply Chain Guru with demand originating at each of the Poké Stops/Gyms to find the shortest means to serve all of them. After a sweet 10 hours, this is the shortest route returned by our solver to cover all 561 Stops:
Figure 1: Screenshot from SupplyChainGuru 8.4 showing the shortest route to cover all stops
Of course you have to use your best judgement to decide which street to walk on given that our routes are point-to-point on the map, not necessarily accounting for the safest route to navigate the city. The route is approximately 35 miles and would take about 17.5 hours of walking under our modeling assumption that a Poké Trainer can walk two miles an hour on an average including the time spent at Poké stops to receive free stuff and catch Pokémon.
While this is a great way to explore all the locations, not every one of us (Yes, even the hardcore Pokémon fans) could afford to spend so much time given all the cool stuff we work on at LLamasoft. What if we only wanted to spend 30 minutes? We added new constraints into the model to determine a route which would allow us to visit the maximum number of locations, if we started from our office located on South Division Street within a 30-minute timeframe. Our solver selected this is as best route:
Figure 2: Best 30 minute route from LLamasoft
Fancy a short Poké walk? This route is about a mile and helps cover 24 Poké stops within 30 minutes. Our next initiative may focus on verifying whether this 30 minute break has any correlation with an increase in productivity among the employees who actively play the game (professional development?). 🙂
With this much success, why stop there? Let’s catch Pokémon every chance we get! This led us to our next scenario, what if we wanted to visit the maximum number of stops while we go out to grab lunch at Chipotle (Making use of the Chiptopia summer rewards program).
This is one of the best things about our software in general, it allows users to model several what-if scenarios based on their unique business needs and compare those results to make the most informed and strategic decision. We now take this route to Chipotle to maximize the benefit of their summer rewards program while managing to catch some Pokémon, all within 30 minutes (Of course, under the assumption that we get through the lines at Chipotle within 10 minutes).
Figure 3: Best Poké route to grab lunch from Chipotle in 30 minutes
Then we decided to mix in some options to multi-task to other popular lunch spots near the office.
Figure 4: Best Poké route to grab lunch from Teriyaki Time in 30 minutes
Figure 5: Best Poké route to grab lunch from Jerusalem Garden in 30 minutes
Figure 6: Best Poké route to grab lunch from Le Dog in 30 minute
Figure 7: Best Poké route to eat lunch at Madras Masala
We’re always excited to learn about the creative ways our customers are leveraging LLamasoft technology and the same applies to employees. Catching Pokémon, getting some exercise during the beautiful Michigan summer and grabbing lunch at one of Ann Arbor’s many great spots: we’re not sure how much more optimal a lunch break could be.
Special thanks to LLamas Brian Dye and Jeremy Castaing for helping out with distances and tweaks to the solver as well as to Jason Judd for his valuable inputs throughout the process.
Want to share more info on the cool ways your team is using LLamasoft technology this summer? Email the Customer Success Team at email@example.com to share your story.