Will people share rides again? Some people never stopped because they don’t have options, putting themselves at great risk. The fact is: traveling, deliveries, and pick-ups have always and will always be an integral part of our daily lives. No matter the purpose of each trip, the logistics of a travel plan is both essential and complicated. However, it’s no surprise that COVID-19 choked out the pre-crisis industry. During the health crisis, close-proximity ride-sharing services like Uber and Lyft suffered drastic 60% to 70% losses globally (McKinsey & Company). Yet, opportunities go largely unfulfilled. Essential services contactless delivery requests for everyday items such as meals, groceries and other essentials skyrocketed as much as by 142% (Chang). Challenging times call for new actions — it’s time to change.

To keep crucial deliveries afloat, six UC Berkeley students joined forces with leading mobility companies, Honda’s 99P Labs and SHARE Mobility. Their goal: develop a scheduled, itinerary fleet optimization model, recommending ideal fleet capacities to cater towards new demands. Unlike ever before, the team tackles novel constraints while doubling down on their competitive edge — fleet advantages like contact-tracing, professional operators and in-advance demand planning. By utilizing digital platforms such as 99PLabs Developer Portal, industry-standard data science practices and ML techniques, the collaboration accelerates optimal scheduling solutions, revolutionizing Mobility-as-a-Service (MaaS) during these critical times.

Data Structures (API)

The key elements of this project follow the traditional data life cycle. First, we worked on learning how to access and pull data from the API, which contained relevant information (trip requests, vehicle information, etc.). Next, the team performed exploratory data analysis (EDA) to further understand the given datasets. This involved basic data cleaning measures as well as the  development of several data visualizations to get a better understanding of each table by analyzing existing correlations and relationships. The team used this for later machine learning models to predict different scenarios and trip request use cases. Next, the team implemented a simple model to assign trips to the nearby vehicles based on minimal distance. Constraints such as number of hours vehicles can be used, specific pickup and dropoff times, time for cleaning vehicles, capacity etc. were incorporated as well. If time permits, we plan to address issues regarding shared rides that involve medicine deliveries between a rider’s pickup and a dropoff location.

System Architecture

As for the final product, team Honda will present our finalized model for trip request optimization and make the appropriate task deployment suggestions to the operations manager of SHARE Mobility for logistical purposes. However, in the future, our team hopes to work on putting together a simplified API that would plug into SHARE Mobility’s backend or even a user interface that presents this information in a more visual, real-time and consolidated manner. By incorporating custom features within the user interface, the operation manager would be able to select different filters, sortings, and restrictions to modify the trip optimization recommendations towards a more catered day-to-day response. With the power of computing, the Honda team ultimately bridges the gap between customers and essential services.

Project by: Olivia Lee, Prosper Amie, Yin Yin Teo, Cindy Li, Misha Lubich, and Chandrima Sabharwal

Github Link: https://github.com/oliviaslee/datax_honda 

Andersson, L., Gläfke, A., Möller, T., & Schneiderbauer, T. (2020, October 19). Why Shared Mobility is Poised to Make a Comeback After the Crisis. Retrieved November 18, 2020, from https://www.mckinsey.com/industries/automotive-and-assembly/our-insights/why-shared-mobility-is-poised-to-make-a-comeback-after-the-crisis

Chang, H; Meyerhoefer, C. D. (2020). COVID ‐19 and the Demand for Online Food Shopping Services: Empirical Evidence from Taiwan. American Journal of Agricultural Economics. doi:10.1111/ajae.12170. Retrieved November 18, 2020, from https://www.nber.org/papers/w27427