What is it about?

Finding compatible passengers in ridesharing is challenging and often left to chance. Existing algorithms prioritize the shortest route, ignoring future requests and traffic, leading to congestion and fewer ridesharing trips. This paper proposes a route recommendation strategy that optimizes demand, reduces congestion, and increases passenger compatibility. Using a multi-task LSTM model, we forecast demand and traffic in city zones to recommend better routes. Tested on datasets from NYC, LA, and Shenzhen, our model achieved 96% accuracy with a 2% RMSE loss. Results show a 23% increase in passenger count for 97% of trips and reduced travel time in 60% of cases. While our approach suggests longer routes in 40% of cases, it enhances driver earnings and reduces passenger wait times.

Featured Image

Why is it important?

Key Findings: - 23% increase in passenger count for 97% of trips - 60% of trips saw reduced travel time despite a longer route in some cases - Improved profitability & sustainability in ridesharing services We tested our approach on datasets from New York City, Los Angeles, and Shenzhen, achieving 96% accuracy in demand prediction! This work has exciting implications for intelligent transportation, route optimization, and sustainable urban mobility.

Perspectives

I very much enjoyed the review process. All the anonymous reviewers gave very constructive feedback on the work, which helped to shape it in its current version. Very much thankful to all the anonymous reviewers, associate editor, and editor in chief of ACM Transactions on Intelligent Systems and Technology.

Jayant Vyas
Indian Institute of Technology Jodhpur

Read the Original

This page is a summary of: PRO-MTL : Parameterized Route Optimization using Multi-Task Learning, ACM Transactions on Intelligent Systems and Technology, February 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3718092.
You can read the full text:

Read

Resources

Contributors

The following have contributed to this page