Summary and Objectives

In recent years, there have been significant advancements in using data-driven techniques to control and coordinate large-scale transportation systems. Typically, these techniques are designed to maximize the efficiency of the system, by minimizing delays and transportation costs. Future transportation systems however should ideally be designed not only to maximize efficiency but also to address societal objectives such as fairness, robustness, privacy, and sustainability. These socially-oriented desiderata introduce several challenges, including (i) requiring the consensus and coordination among agents on acceptable definitions of and tradeoffs between these desiderata, (ii) the computational intractability of traditional approaches, and (iii) the consideration of data-availability issues due to privacy concerns.

Learning-based approaches have the potential to address these issues and help extend the capabilities of traditional control algorithms for transportation applications; however, the use of data-driven techniques for the societally aware operation of emerging mobility systems is largely untapped. The objectives of the workshop are four-fold:

• Emphasize data-driven machine learning techniques in the context of transportation applications. The discussion of data-driven modeling, analysis, and control tools in applications such as road traffic management will help foster methodological developments to design societally aware algorithmic decision-making systems for emerging mobility systems.

• Provide a platform to explore new research directions geared towards tapping the full potential of data in achieving societal objectives in transportation systems.

• Foster the development of novel control mechanisms essential for advancing societally aware transportation systems and enable discussions on the appropriate measures for societal considerations such as fairness, equity, robustness, privacy etc.

• Introduce the controls community to transportation-specific models for societal objectives, highlight limitations of the current models, and challenges faced in implementing these objectives using traditional approaches. 

Expected Outcomes

We expect three main outcomes for participants of this workshop:

1. The engagement of researchers across control theory, operations research, economics, computer science, and transportation-related fields in the discussion of recent advances in the design of learning based systems for the control and coordination of transportation systems. 

2. Fostering new connections and an exchange of ideas across researchers from diverse backgrounds within the context of emerging mobility systems while enabling young and established researchers alike to connect on areas of related interest.

3. Summarizing open problems in this area, with the hope of equipping participants with the necessary tools and literature to pursue this novel, and societally relevant research area.

Target Audience

The target audience of the workshop comprises graduate students, academic researchers, and industry experts with interest in using learning based approaches for the societally-aware control and coordination of future transportation systems. Furthermore, we welcome all researchers and practitioners in the fields of control theory, operations research, computer science, economics, and transportation to explore new frontiers.

Organizers

Alexandre Bayen (UC Berkeley)

Karthik Gopalakrishnan (Stanford University)

Devansh Jalota (Stanford University)

Jessica Lazarus (UC Berkeley)

Marco Pavone (Stanford University)