Overview
RPM eco offers an integrated management system for recycling of contaminated plastic hydrocarbon containers, as well as pesticide and fertilizer containers. The company has thousands of customers all across Canada.
SimWell is a consulting company with a global team of engineers and simulation consultants dedicated to simulation, optimization, and digital twins. They help trailblazing business leaders make confident decisions with the use of simulation modeling.
Problem
RPM eco drivers picked up the recyclables from their customers and brought them back to the collection center. One of the main issues was that all those clients were on a fixed time interval. This meant that they would be serviced once a week, once every two weeks, or once a month.
Another problem was that all of those customers were on a static route. This could be an issue if there was seasonality in consumption. Only two employees managed all the service calls and operations, such as building these routes.
RPM eco had three main stakeholders: drivers, coordinators, and managers.
Drivers were paid by the amount of weight they brought back to the collection center. They would like to have more kilograms per kilometer done, much fewer empty pickups, and an easy-to-use mobile app.
Coordinators wanted more visibility on the ongoing routes and their drivers. Also, they intended to increase the drivers’ willingness to operate the routes.
As for managers, it was necessary to increase customer service levels and employee satisfaction.
RPM eco planned to get to their customers right before they called to maximize the weight picked up and have a high service level for their customers. RPM eco also wanted to build dynamic routes according to the needs of their customers and be able to easily modify these routes and keep track of them. The business purpose was to maximize the material weight collected by trucks.
Solution
SimWell helped RPM eco optimize a reverse logistics supply chain. For this purpose, SimWell developed their own end-to-end solution called OSCAL. It consisted of prediction, simulation, and a mobile app. In the picture below, OSCAL is illustrated.
Prediction of the available weight for each customer
This end-to-end solution started with prediction. SimWell engineers created a predictive tool to anticipate the weight available in each of the client’s facilities. Predictive algorithms were based on historical data and seasonality. The engineers used two methods: ARIMA (autoregressive integrated moving average) and Croston, mostly used for intermittent demand.
They combined the predictive tool with a supervised learning model to manage client variability. They were able to anticipate if the client’s consumption behavior required them to change from one method to another.
After SimWell implemented these tools, RPM eco was able to increase the service level and reduce the margin of error between the predicted and pickup weights. They could predict the available weight for each customer at a certain point in time.
Simulation of routes for optimizing the travel distance between customers
Previously, RPM eco had static routes that were problematic due to seasonality. To address these challenges, SimWell developers built an agent-based simulation model to manage the prioritization of routes. They used mostly statecharts for modeling the processes. Usually, in the normal process, RPM eco would use it week by week to get the best predictions for the customers.
The simulation model received the data from a cloud database. The model developers launched the Python scripts to readjust the intervals for the new customers and then do the predictions. After these predictions, engineers had a list of available clients for the simulation.
The modelers had different route creation algorithms using specific prioritization and GIS. This allowed engineers to keep track of driving time, which is important for the trucking industry. They wanted to optimize the travel distance between the customers on this route. Once the route was accepted or rejected, the information was sent back to the cloud database.
There were tools that allowed the coordinators to work with these routes. On the left side of the picture below, there is a tool called Routes Planner that enables specialists to add service calls, modify routes, or remove clients in case of changes. On the right side of the picture below, there is a Customer Care Center that is used to modify these clients according to their needs.
Development of a mobile app for drivers
As soon as the engineers predicted the volumes and created the routes, the drivers could receive their routes. Previously, they received information on paper, but this was inefficient and unreliable because it could be lost in transit. Additionally, this paper element made it difficult to implement this simulation model. Considering these facts, the SimWell specialists decided to develop a mobile application.
The application could handle the inputs and outputs of the simulation automatically. The driver could complete the route once he logged in to the app. Then the information about this was sent back to the simulation model.
The database was based on MySQL and hosted on the Azure portal. This handled the operational work of RPM and acted as a communication medium between the app and the simulation.
The app also provided validation on the data pickups, so it had a predicted weight and threshold. If the data was outside the threshold, then the driver would be asked to validate and edit if necessary. This app closed the loop on the end-to-end solution.
Infrastructure of the solution
The solution infrastructure consisted of several elements: the Microsoft Azure SQL database, Power BI, Power Apps, the predictions, and the mobile app. Power BI and Power Apps were connected to Microsoft Azure for data visualization and modifications of parameters. On top of that, the AnyLogic simulation model simply coordinated all the elements.
GIS was already integrated into AnyLogic, which was important for this solution. As reverse logistics software, AnyLogic enabled engineers to easily integrate and exploit the external Azure SQL database as well as Python scripts for predictive analytics in logistics.
Results
All the stakeholders benefited from this solution. The coordinator had a bird’s-eye view in real time of the reverse logistics supply chain. The BI tools that were integrated helped operations and management teams make tactical and strategic decisions. Overall, there was much better operational data with higher security and governance.
Since the beginning of the project, RPM eco has increased the total weight collected from customers by 20% and the average weight per pickup by 52%. At the same time, the number of empty pickups has decreased by 81%.
The first chart in the illustration below shows the tremendous decrease in zero pickups from 2021 to 2022. The second chart demonstrates the increase in the average weight by pickup.
The case study was presented by Shirwa Mahdi, Maxime Leclerc, Joseph Moyer, David Larivee, of SimWell, at the AnyLogic Conference 2022.
The slides are available as a PDF.
