Deploying Mobility-On-Demand for All by Optimizing Paratransit Services

Sophie Pavia, David Rogers, Amutheezan Sivagnanam, Michael Wilbur, Danushka Edirimanna, Youngseo Kim, Philip Pugliese, Samitha Samaranayake, Aron Laszka, Ayan Mukhopadhyay, Abhishek Dubey

Abstract: While on-demand ride-sharing services have become popular in recent years, traditional on-demand transit services cannot be used by everyone, e.g., people who use wheelchairs. Paratransit services, operated by public transit agencies, are a critical infrastructure that offers door-to-door transportation assistance for individuals who face challenges in using standard transit routes. However, with declining ridership and mounting financial pressure, public transit agencies in the USA struggle to operate existing services. We collaborate with a public transit agency from the southern USA, highlight the specific nuances of paratransit optimization, and present a vehicle routing problem formulation for optimizing paratransit. We validate our approach using real-world data from the transit agency, present results from an actual pilot deployment of the proposed approach in the city, and show how the proposed approach comprehensively outperforms existing approaches used by the transit agency. To the best of our knowledge, this work presents one of the first examples of using open-source algorithmic approaches for paratransit optimization.

Introduction

Paratransit serves as a vital lifeline for individuals with disabilities who are unable to utilize standard fixed-route public transportation. Despite its necessity, these services are often plagued by extreme operational inefficiencies. In Chattanooga, Tennessee, for instance, paratransit services managed by the Chattanooga Area Regional Transportation Authority (CARTA) account for 22% of total service miles while serving less than 4% of total passenger trips. This research presents an open-source, data-driven optimization framework designed to address these inefficiencies through advanced vehicle routing and human-centric design.

The Complexity of Paratransit Optimization

Optimizing paratransit is significantly more complex than standard ride-sharing for several reasons:

  1. Regulatory Compliance: The Americans with Disabilities Act (ADA) requires paratransit to be comparable to fixed-route services, necessitating strict adherence to pickup and drop-off time windows.
  2. Variable Vehicle Capacity: Unlike standard cars, paratransit vehicles feature flexible seating. A single wheelchair occupies the space of multiple ambulatory seats, and vehicle configurations must often be adjusted in real-time.
  3. Tacit Knowledge and Social Factors: Dispatchers often rely on unwritten knowledge, such as specific passenger-driver rapport, which traditional "black-box" algorithms fail to capture.

Methodology: A Human-in-the-Loop Framework

The researchers formulated the challenge as a Vehicle Routing Problem with Pickup and Delivery and Time Windows (VRPPDTW). The solution employs a two-tiered algorithmic approach:

  • Offline Optimization: For advance bookings, the system uses Guided Local Search (via Google’s OR-Tools) to minimize total vehicle miles traveled (VMT).
  • Online Optimization: For same-day requests, an insertion heuristic slots new trips into existing manifests by minimizing the ratio of VMT to Passenger Miles Traveled (PMT).

Crucially, the framework includes a "Human-in-the-Loop" web interface. This allows supervisors to manually adjust schedules via drag-and-drop while a built-in constraint checker ensures that ADA timing and vehicle capacity limits are never violated.

The paratransit scheduling interface allows for manual edits with real-time constraint validation.

Data and Experimental Setup

The study utilized comprehensive operational data from CARTA (August 2023), including daily request logs ranging from 125 to 159 trips. These logs detailed GPS coordinates and specific capacity needs, distinguishing between ambulatory and wheelchair-bound passengers. The road network was modeled using OpenStreetMap (OSM) to ensure accurate travel time matrices.

Daily paratransit ridership data highlighting the mix of ambulatory and wheelchair users.

Key Results and Pilot Success

The proposed system was validated through both simulations and a live two-day pilot deployment in Chattanooga. The results demonstrated significant improvements over legacy commercial software:

  • Efficiency Gains: The offline solver reduced the VMT/PMT ratio by 21.3%. During the live pilot, the system saved 356 miles on the first day and 236 miles on the second—a 17–24% efficiency boost.
  • Improved Shared Rates: The "shared rate" (the percentage of trips with multiple passengers) jumped from 53% in the baseline to 85% with the new solver.
  • Operational Reliability: While legacy systems frequently produced schedules with time-window violations, the new open-source approach eliminated these violations entirely.
  • Administrative Speed: Scheduling tasks that previously required six hours of manual labor were completed in minutes.

Conclusion and Future Directions

This work demonstrates that open-source algorithmic frameworks can significantly outperform proprietary legacy systems in public transit. While the current system excels with advance bookings, future research will focus on improving the online solver's performance for scenarios where real-time requests exceed 40% of total volume. By bridging the gap between mathematical optimality and the practical needs of dispatchers, this framework provides a scalable model for more inclusive and efficient urban mobility.

Summary updated

Abstract

While on-demand ride-sharing services have become popular in recent years, traditional on-demand transit services cannot be used by everyone, e.g., people who use wheelchairs. Paratransit services, operated by public transit agencies, are a critical infrastructure that offers door-to-door transportation assistance for individuals who face challenges in using standard transit routes. However, with declining ridership and mounting financial pressure, public transit agencies in the USA struggle to operate existing services. We collaborate with a public transit agency from the southern USA, highlight the specific nuances of paratransit optimization, and present a vehicle routing problem formulation for optimizing paratransit. We validate our approach using real-world data from the transit agency, present results from an actual pilot deployment of the proposed approach in the city, and show how the proposed approach comprehensively outperforms existing approaches used by the transit agency. To the best of our knowledge, this work presents one of the first examples of using open-source algorithmic approaches for paratransit optimization.
BibTeX
@inproceedings{pavia2024deploying,
  author = {Sophie Pavia and David Rogers and Amutheezan Sivagnanam and Michael Wilbur and Danushka Edirimanna and Youngseo Kim and Philip Pugliese and Samitha Samaranayake and Aron Laszka and Ayan Mukhopadhyay and Abhishek Dubey},
  title = {{Deploying Mobility-On-Demand for All by Optimizing Paratransit Services}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {2024},
  url = {https://openreview.net/forum?id=mS9g0JXD2l},
}
Figure