Stochastic and Learning-Based Control Strategies for Electric Autonomous Mobility Systems
Doktorsavhandling, 2026
Electric Autonomous Mobility-on-Demand (E-AMoD) systems offer a path toward sustainable urban transportation through the coordinated operation of shared, zero-emission autonomous vehicles. Yet their deployment poses difficult operational challenges: fleet rebalancing, vehicle routing, and charging must be managed jointly, under uncertainty, and within tight computational budgets. A central argument of this thesis is that no single decision-making paradigm suffices. Optimization-based methods are well suited for strategic fleet control where uncertainty guarantees and feedback are essential, while learning-based methods become necessary at finer operational scales where real-time optimization is computationally prohibitive.
Three contributions are presented, each targeting a different operational level. The first introduces a chance-constrained model predictive control (MPC) framework for station-level fleet rebalancing, combining Gaussian Process Regression for probabilistic demand forecasting with a hierarchical architecture that separates strategic rebalancing from tactical matching. The second extends this framework to electric fleets operating under multiple interacting uncertainties, employing a tailored Nested Benders Decomposition to maintain metropolitan-scale tractability without sacrificing MPC's receding-horizon feedback. The third contribution shifts to node-level electric dial-a-ride routing, including pickup-delivery sequencing, time windows, and ride-time constraints, and proposes a deep reinforcement learning approach built on a Graph Edge Attention Network capable of handling hundreds of requests with second inference times. Taken together, the three contributions show that optimization and learning serve complementary roles in E-AMoD operations, with the appropriate paradigm determined by the granularity and real-time demands of the problem at hand.
Three contributions are presented, each targeting a different operational level. The first introduces a chance-constrained model predictive control (MPC) framework for station-level fleet rebalancing, combining Gaussian Process Regression for probabilistic demand forecasting with a hierarchical architecture that separates strategic rebalancing from tactical matching. The second extends this framework to electric fleets operating under multiple interacting uncertainties, employing a tailored Nested Benders Decomposition to maintain metropolitan-scale tractability without sacrificing MPC's receding-horizon feedback. The third contribution shifts to node-level electric dial-a-ride routing, including pickup-delivery sequencing, time windows, and ride-time constraints, and proposes a deep reinforcement learning approach built on a Graph Edge Attention Network capable of handling hundreds of requests with second inference times. Taken together, the three contributions show that optimization and learning serve complementary roles in E-AMoD operations, with the appropriate paradigm determined by the granularity and real-time demands of the problem at hand.
Model Predictive Control
Optimization Under Uncertainty
Mobility-on-Demand
Electric Autonomous Vehicles
Stochastic Programming
Graph Neural Networks.
Benders Decomposition
Robust Optimization
Deep Reinforcement Learning
SB-H6, Sven Hultins gata 8, Göteborg
Opponent: Prof. Carolina Osorio, HEC Montréal, Canada
Författare
Sten Elling Tingstad Jacobsen
Chalmers, Elektroteknik, System- och reglerteknik
A Predictive Chance Constraint Rebalancing Approach to Mobility-on-Demand Services
Communications in Transportation Research,;Vol. 3(2023)
Artikel i vetenskaplig tidskrift
Imagine ordering a ride from your phone and a self-driving electric car arrives within minutes, picks you up, and drops you off without a human driver. Now imagine thousands of these vehicles operating across a city simultaneously, sharing trips, recharging intelligently, and repositioning themselves to be where passengers will need them next. This is the vision of Electric Autonomous Mobility-on-Demand: a shared transportation system that combines electric vehicles, autonomous driving, and on-demand service to offer clean, affordable, and accessible mobility at scale.
But running such a fleet efficiently is an enormously complex problem. Passengers arrive unpredictably, traffic varies throughout the day, and vehicles need to recharge before their batteries run out. Every decision about where to send a vehicle now affects what options are available minutes later, and the future is always uncertain. This thesis develops methods that help a fleet of electric autonomous vehicles make smarter decisions despite that uncertainty.
The first contribution addresses how a fleet should reposition empty vehicles to be ready for future passengers. Rather than reacting to requests as they arrive, the system uses demand forecasts with uncertainty estimates to provide a concrete guarantee: there is at least a 95 percent chance that enough vehicles will be in the right place at the right time.
The second contribution tackles the same problem for electric fleets, where vehicles must also decide when and where to charge. The method handles uncertain demand, unpredictable travel times, and varying energy consumption simultaneously, and a decomposition algorithm solves the resulting large optimization problem in parallel, reducing computation from hours to minutes.
The third contribution addresses the matching problem: when a passenger requests a ride, which vehicle should serve them, and in what order should multiple passengers be picked up and dropped off? An artificial intelligence agent trained on millions of simulated trips learns to find high-quality solutions in seconds, fast enough to respond to live requests.
Together, these contributions show that uncertainty-aware decision-making can make electric autonomous fleets both more reliable and more efficient, bringing the vision of clean, shared urban mobility closer to reality.
But running such a fleet efficiently is an enormously complex problem. Passengers arrive unpredictably, traffic varies throughout the day, and vehicles need to recharge before their batteries run out. Every decision about where to send a vehicle now affects what options are available minutes later, and the future is always uncertain. This thesis develops methods that help a fleet of electric autonomous vehicles make smarter decisions despite that uncertainty.
The first contribution addresses how a fleet should reposition empty vehicles to be ready for future passengers. Rather than reacting to requests as they arrive, the system uses demand forecasts with uncertainty estimates to provide a concrete guarantee: there is at least a 95 percent chance that enough vehicles will be in the right place at the right time.
The second contribution tackles the same problem for electric fleets, where vehicles must also decide when and where to charge. The method handles uncertain demand, unpredictable travel times, and varying energy consumption simultaneously, and a decomposition algorithm solves the resulting large optimization problem in parallel, reducing computation from hours to minutes.
The third contribution addresses the matching problem: when a passenger requests a ride, which vehicle should serve them, and in what order should multiple passengers be picked up and dropped off? An artificial intelligence agent trained on millions of simulated trips learns to find high-quality solutions in seconds, fast enough to respond to live requests.
Together, these contributions show that uncertainty-aware decision-making can make electric autonomous fleets both more reliable and more efficient, bringing the vision of clean, shared urban mobility closer to reality.
TSIM: Simulering, analys och modellering av framtida effektiva trafiksystem
VINNOVA (2018-05003), 2019-09-01 -- 2023-09-01.
Styrkeområden
Transport
Ämneskategorier (SSIF 2025)
Reglerteknik
DOI
10.63959/chalmers.dt/5841
ISBN
978-91-8103-384-7
Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie: 5841
Utgivare
Chalmers