Join us to participate in the upcoming 2019 City:One Challenge. 


T-REX is a novel crowd-sourced decision-support system for improving public transit riding experience.

Photo of Scott Averitt
1 1

Written by

In this project we propose T-REX, a novel crowd-sourced decision-support system for improving public transit riding experience. Realized as a mobile app, T-REX aims to integrate past and current transit data (e.g., locations of the buses/trains, road closures and blocks), both publicly available and rider-generated ones, with rider intent information, to predict travel time, comfort, and cost of various transit options. Furthermore, our proposed system builds behavior models, decision histories, and assesses the impact of the rider’s decisions on public resources and on other riders over time, encouraging collaboration, hence overall better transit experience.

Public transportation is a common good, and riders have collaborative and competitive relationships when consuming it. Selfish behavior negatively affects the community: for instance, if a rider takes public transportation only based on her own schedule and preferences, this might lead to over-crowded buses/trains/subways during rush hour. In contrast, a solution that allows every rider to plan trips by considering other riders’ behavior and preferences, along with context-relevant information, would facilitate the distribution of public transit resources, resulting in an improved experience for the entire community.  

To augment customer satisfaction, a smart solution should be able to unpack both public and private dimensions of rider experience and promote collaborative behavior by balancing them. Public dimensions include events that affect public transportation's arrival and travel times, such as emergencies, public gatherings like concerts, marathons, tailgate parties, protests, etc. Private dimensions include reasons for travel (e.g., daily commute, leisure), distance between origin and destination, willingness to wait with or without incentives (e.g., less crowded buses outside rush hour, discounts), alternative route choices, and personal preferences.

In this regard, a decision support system that aggregates public and private dimensions of rider experience at scale, processing them using Artificial Intelligence can be used to augment riders’ context awareness, thus customizing the transit solution and, ultimately, mitigating riders’ frustration. T-REX aims to be such a decision-support system for public riders in the Pittsburgh region. The system, implemented as a mobile app, will support urban and suburban routes of Port Authority bus, light rail, and subway systems. T-REX will improve customer experience by analyzing local riders’ competition on public transit resources and encouraging collaboration. T-REX will allow local riders to share information collaboratively to decide better timing and routing of taking public transit, in both self and public interests. Our system will also mediate the competing relationships when consuming public transportation resources by influencing individuals’ decision-making related to transit timing and routes: we will model the riders’ behavior as evolutionary game with additional information of past interactions of each other and present the equilibrium (best transit option in response to other riders’ decisions) to the riders. More importantly, via incentive design and novel user interfaces, we will also show riders alternative transit options that are not “the best” options but much better in terms of both self and public interest.

The mobile app will also show how transit alternatives affects other riders who have decided not to compete with the user for the transit resource, and how likely other riders will help the user in the future. By providing such options through a user-friendly interface, riders can easily interpret the impact of their decisions: in this regard, they are encouraged to collaborate and push the equilibrium to benefit the public transportation system as a whole. T-REX aims to integrate past and current transit data (e.g., locations of the buses/trains, road closures and road conditions), both publicly available and rider-generated ones, with the current intent of other T-REX users to predict the travel time, comfort, and cost of various transit options.

Such options include both timing and routes of transit. Furthermore, T-REX builds behavior models, decision histories, and assesses impact of the rider’s decisions on public resource and other riders’ experience over time. When presenting transit options to the riders, the impact on riders who have taken altruistic alternatives, as well as the displaying potential future benefit from other riders for each transit option.

Past approaches on decision-making do not consider the “potential riders”, i.e., riders that are just about to decide whether/how they are going to use public transport, but only the riders in the system currently and status of the transit system. This implies that state of the art tools tend to present the same information and suggest the same decisions to all potential riders, which pushes them to make similar decisions ending up with more competition on public resources. T-REX will take each potential rider’s interest and flexibility into account, and suggest different alternatives that fit each individual’s schedule to encourage riders making community-aware decisions.

Describe who will use your solution (1,000 characters)

T-REX aims to improve the life of Pittsburghers and, more in general, to make the experience of people moving across urban and suburban areas of the Steel City better. Unlike state of the art mobility solutions, T-REX is not only directed to “power users” of public transportation (e.g., everyday commuters): customer profiles can be built and updated in real-time by fetching and aggregating information from social networks, so that preferences and intents of casual riders can be seamlessly correlated to the current mobility offer, even without a prior public transportation history. T-REX recommendations based on AI: in particular, semantic technologies are used for aggregation and representation of user profiles, and machine-learning algorithms are used for real-time correlation between profile features and transportation solutions.

Describe your solution's stage of development

  • Initial Design - you are still exploring the idea and have not tested it with users
  • Prototype - you have built a prototype and tested it with potential users

Tell us about your team or organization (500 characters)

The Bosch Group employs roughly 402,000 associates worldwide and generated global sales of $88.2 billion (as of December 31, 2017). Its operations are divided into four business sectors: Mobility Solutions, Industrial Technology, Consumer Goods, and Energy and Building Technology. Having established a presence in North America in 1906, Bosch employs nearly 32,800 associates in more than 100 locations. In 2017, Bosch generated consolidated sales of $13.7 billion in North America.

Size of your team or organization

  • 501+

Funding Request

  • $25,000

Describe how you would pilot your idea (1000 characters)

Bosch Corporate Research will build a prototype system in 3 months, by combining the expertise of three research scientists from our Pittsburgh team with a mobile app developer hired by taking $ 15,000 from the prize. We will then use the remaining $10,000 to fund a beta-test with a limited number of users, suitably selected across social groups. Funding the beta-test includes incentives and cost of server hosting. After the pilot is completed, and the effectiveness of the solution is demonstrated, we will collaborate with the Bosch Smart City business unit to build a cloud service around the T-REX idea. The service will be initially free to riders for a weeklong trial period, and subsequently available for a monthly subscription fee.

Describe how you would measure the success of your pilot (1000 characters)

Success will be evaluated by combining mobile app acquisition and user satisfaction. The very first challenge in developing our idea beyond the pilot phase is to motivate riders in using the T-REX mobile app, and quickly build up a critical mass: without a large-scale participation, which is a requirement for crowd-based systems, optimal recommendations cannot be generated. In order to tackle this challenge, we plan to collaborate with local stakeholders and build a system of incentives customized to Pittsburgh residents (access to local promotions and discounts, free rides, etc.). Such a mechanisms of incentive design will support app acquisition and retention. Achieving large-scale adoption in only possible if the quality of the user experience and engagement is high. In this regard, we plan to combine 1) a simple star-rating method to assess user feedback following a T-REX recommendation, with 2) thorough follow-up interviews of beta-testers.

Attachments (1)

1 comment

Join the conversation:

Photo of Diana Avart

Hi Scott,

Thanks so much for submitting this idea. It sounds like you and the Bosch team have really thought through this idea, from the technology, to the team that would work on it, to the incentives for the users.

- Diana, Facilitator