Reinventing Urban Mobility: On-Demand, Optimized, and Efficient

Javier Alon Somura and his team are redefining how cities move. By combining autonomous vehicles with ride-hailing and advanced motion-planning algorithms, they’ve developed high-capacity ride pooling capable of serving hundreds of thousands of trips daily—while keeping vehicles full and traffic manageable. Their methods, proven in large-scale Manhattan simulations and deployed commercially worldwide, are now extending to flash deliveries, warehouse logistics, and even integrated public transit, showing how smart planning can make urban mobility faster, safer, and greener.

Transcript

00:00:01 Hello, I'm Javier Alon Somura and it's my pleasure to give you a brief overview of our work in on demand transportation. About 10 years ago, I got very interested in this challenge of having efficient and safe mobility for all. Towards this goal, we have contributed methods for motion planning in urban environments. in particular modeling the interaction with other decision making

00:00:27 agents such as bikers, pedestrians or other vehicles. Our methods have addressed multimodel uncertainty in predictions. We have also modeled the complex interaction with other traffic participants for example by using social value orientation interaction aware and PPI or by formulating contingency games. We have also provided tools for global guidance

00:00:55 of local planners like in this example to merge and create by creating a gap into a lane full of other vehicles. Autonomous vehicles are being combined with right hailing to provide efficient and safe mobility for all. However, they may not make our cities greener. There are many studies that have shown that right hailing can in fact lead to a large increase in

00:01:22 traffic. A solution for this is to have on demand right pooling where multiple people with similar origin and destination can be combined in a vehicle at the same time. We have proposed the first optimizationbased method capable of large scale highcapacity right pooling. We demonstrated our method with extensive simulations of Manhattan with

00:01:50 fleets of thousands of vehicles servicing about half a million requests per day. We show that our method can achieve high occupancy of the vehicles while remaining realtime capable. A company the routing company now commercialized these methods and has served over 3 million passengers in more than 50 deployments worldwide.
>> [snorts]

00:02:15
>> We have also proposed many extensions for distributed task assignment for flash deliveries and warehouse logistics for multi-objective optimization and fleet design as well as to include walking predictive routing and equilibrium analysis. In this regard, our latest work has been in combining the flexible ondemand vehicles with fixed bus lines to have an

00:02:44 integrated public transit. With this, I would like to thank uh students, collaborators and colleagues as well as our funding sources. Without all of you, this will not be possible. Thank you very much.