Assistant Professor in Smart Transportation

DiDi Chuxing Professor

School of Vehicle and Mobility

Tsinghua University

Email: cwangkai@tsinghua.edu.cn

1. This paper optimizes routes with flexible pickups and drop-offs. This feature is commonly found in ride-sharing (Uber Express Pool) and smart mobility systems: instead of meeting at pre-specified origins and destinations, customers and vehicles can meet at more convenient (optimized) locations.

2. This paper develops a strong model formulation based on network flows and an efficient decomposition algorithm. To support ride-sharing services at a large scale in real-time, it also designs an online optimization framework, which is tested in a real-world setting of New York City. 

3. Results show that the algorithm can solve problems with up to 200 customers in less than one minute offline, and can successfully tackle the New York  Taxi system implementation online. Practically, floating targets increase 8% profit, and flexible pickups are more beneficial than flexible drop-offs. 

1. Electric vertical-takeoff-and-landing vehicles enable a new form of mobility known as Urban Aerial Mobility (UAM) and use vertiports to takeoff and land. This paper optimizes the number, location and capacity of vertiports in a metropolitan area—the key strategical planning problem for UAM.

2.  This paper formulates a model to capture inter-dependencies between vertiport deployment,   UAM operations and customer adoption, including a  "tractable part" but also an intractable non-convex customer adoption function.  It develops an original exact algorithm based on adaptive discretization.

3. The algorithm converges to a 1% optimality gap. Practically, UAM networks vary widely across metropolitan areas, as a function of geographic, urban and commuting patterns. Vertiport networks grow in a nested fashion, starting with a few ``obvious" vertiports and adding vertiports as penetration increases.   Airport shuttle and long-distance commutes are two main potential use cases.

1.  This paper proposes an integrated model of scheduling and operations in airport networks that jointly optimizes scheduling interventions and ground-holding operations across airports networks under operating uncertainty.

2. The problem is formulated as a stochastic integer program. To solve it, the paper develops an original decomposition algorithm, based on new optimality cuts—dual integer cuts—that leverage the reduced costs of the dual linear programming relaxation of the second-stage problem.

3. Computational experiments show that the algorithm yields near-optimal solutions for the entire U.S. National Airspace System network. Ultimately, the proposed approach enhances airport demand management models through scale integration and scope integration.

1. This paper addresses a problem of two-stage data-driven optimization under binary uncertainty, a widely prevalent class of problems, e.g., a two-stage problem with first-stage planning and second-stage routing decisions, under uncertainty regarding which customers to visit.

2. It first defines a scenario-based robust optimization (ScRO)   that combines principles of stochastic programming and robust optimization. Second, it proposes deviation likelihood and sample clustering methods to generate scenarios and uncertainty sets from element-wise probabilities or historical samples. Third, it develops a sparse row generation algorithm to solve the ScRO. 

3. Using public datasets, results suggest that (i) the ScRO formulation outperforms benchmarks based on deterministic, stochastic and robust optimization; (ii) the deviation likelihood method outperforms scenario generation baselines; and (iii) the sparse row generation algorithm returns stronger solutions in shorter computational times than off-the-shelf implementation.