Yixin Zhu
Research Subjects
Yixin Zhu has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- Jian Sun
Tongji University
3 shared publications Yiru Liu
3 shared publicationsZirui Li
3 shared publications- Haotian Shi
Department of Civil and Environmental Engineering, University of Wisconsin-Madison, Madison, WI, USA
2 shared publications - Xiaocong Zhao
Tongji University
2 shared publications Haotian Shi
1 shared publication
Latest Publications
- 2 November 2026
- Published by Elsevier in Accident Analysis & Prevention
- Vol. 240
Why some emergency maneuvers succeed: Interpretable behavior graphs from expert drivers to automated control
- 2 November 2026
- Published by Elsevier in Accident Analysis & Prevention
- Vol. 240
Why some emergency maneuvers succeed: Interpretable behavior graphs from expert drivers to automated control
- Open Access
- 16 October 2026
- Published by Elsevier in SSRN Electronic Journal
Does Intelligent Travel Decision-Making Improve Traffic System Efficiency? Evidence from Dual-Timescale Incentive-Based Travel Demand Management
With advances in artificial intelligence and automated driving, travelers may increasingly delegate their travel decisions to autonomous agents. Compared with experience-based human adjustment, autonomous agents can continuously revise their travel decisions in response to traffic conditions. These different response mechanisms may produce different day-to-day traffic patterns, raising the question of whether more intelligent individual decision-making capability improves overall system performance. To examine this issue, this study develops a dual-timescale incentive-based travel demand management (IBTDM) framework that jointly optimizes within-day time varying incentives and day-to-day budget allocation under a fixed budget to mitigate congestion. Building on the bottleneck model, we first represent the day-to-day evolution of human travelers’ departure time choices using the proportional switch adjustment process and solve the resulting incentive design problem with a physics driven differentiable optimization framework. We then consider autonomous traveler decision-making, in which traveler agents learn their departure time strategies using a stateless Q-learning algorithm, while the administrator adjusts the daily incentive profile based on the aggregate system state. The results show that, at the same budget level and operational period, human decision-making leads to a lower average daily system travel cost than autonomous decision-making. Additionally, for a given operational period, the optimal budget is higher for autonomous agents than for human travelers. These findings demonstrate that more intelligent individual decision-making capability does not necessarily improve system performance, as strategic interactions among autonomous agents make incentive schemes less effective. They also provide practical guidance for designing incentive schemes under different traveler decision-making mechanisms.
- 16 October 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessDoes Intelligent Travel Decision-Making Improve Traffic System Efficiency? Evidence from Dual-Timescale Incentive-Based Travel Demand Management
With advances in artificial intelligence and automated driving, travelers may increasingly delegate their travel decisions to autonomous agents. Compared with experience-based human adjustment, autonomous agents can continuously revise their travel decisions in response to traffic conditions. These different response mechanisms may produce different day-to-day traffic patterns, raising the question of whether more intelligent individual decision-making capability improves overall system performance. To examine this issue, this study develops a dual-timescale incentive-based travel demand management (IBTDM) framework that jointly optimizes within-day time varying incentives and day-to-day budget allocation under a fixed budget to mitigate congestion. Building on the bottleneck model, we first represent the day-to-day evolution of human travelers’ departure time choices using the proportional switch adjustment process and solve the resulting incentive design problem with a physics driven differentiable optimization framework. We then consider autonomous traveler decision-making, in which traveler agents learn their departure time strategies using a stateless Q-learning algorithm, while the administrator adjusts the daily incentive profile based on the aggregate system state. The results show that, at the same budget level and operational period, human decision-making leads to a lower average daily system travel cost than autonomous decision-making. Additionally, for a given operational period, the optimal budget is higher for autonomous agents than for human travelers. These findings demonstrate that more intelligent individual decision-making capability does not necessarily improve system performance, as strategic interactions among autonomous agents make incentive schemes less effective. They also provide practical guidance for designing incentive schemes under different traveler decision-making mechanisms.
- Open Access
- 15 July 2026
- Published by Elsevier in SSRN Electronic Journal
Why Some Emergency Maneuvers Succeed: Interpretable Behavior Graphs from Expert Drivers to Automated Control
Emergency handling is among the hardest capabilities for automated driving. Safe evasion demands staged, anticipatory, and tightly coupled longitudinal--lateral actions, yet the expert logic behind such maneuvers is scarce, hard to label by skill, and locked in continuous control signals that obscure why some maneuvers succeed and others fail. This study introduces the Driving Behavior Graph (DBG), an interpretable, probabilistic representation that turns scarce expert demonstrations into reusable knowledge linking human expert decision logic to machine-executable control. Through a two-stage procedure, we screen 25 expert drivers from 99 candidates and collect 400 driver-in-the-loop trials across four emergency scenarios (jaywalking, cut-out, cut-in, merge-in) on a virtual-reality platform. Control signals are segmented with a Hierarchical Dirichlet Process Hidden Semi-Markov Model, mapped to a scenario-specific semantic grid, and assembled into a probabilistic DBG of behavior chains and transitions. The DBG shows that expert emergency behavior is proactive and phased: evasive action is initiated far earlier (lateral-maneuver onset at about 55 vs. 115 m in the cut-out scenario), longitudinal and lateral control are coordinated, and the vehicle is deliberately recovered after the conflict; it further localizes the bifurcation points where excellent and poor episodes diverge. Notably, self-assessed driving skill does not predict objective emergency performance, indicating that competence in safety-critical handling cannot be inferred from routine self-report. To close the loop, DBG-derived primitives are synthesized into reference trajectories that guide a model predictive controller, which reproduces expert-like earlier lateral evasion (p < 0.001) and a deliberate safety-over-compliance trade-off, and runs in a vehicle-in-the-loop deployment on a Level-3 automated test vehicle. To support reproducibility, the skill-screened expert emergency-driving dataset is publicly available at https://zenodo.org/records/20710939.
- 15 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessWhy Some Emergency Maneuvers Succeed: Interpretable Behavior Graphs from Expert Drivers to Automated Control
Emergency handling is among the hardest capabilities for automated driving. Safe evasion demands staged, anticipatory, and tightly coupled longitudinal--lateral actions, yet the expert logic behind such maneuvers is scarce, hard to label by skill, and locked in continuous control signals that obscure why some maneuvers succeed and others fail. This study introduces the Driving Behavior Graph (DBG), an interpretable, probabilistic representation that turns scarce expert demonstrations into reusable knowledge linking human expert decision logic to machine-executable control. Through a two-stage procedure, we screen 25 expert drivers from 99 candidates and collect 400 driver-in-the-loop trials across four emergency scenarios (jaywalking, cut-out, cut-in, merge-in) on a virtual-reality platform. Control signals are segmented with a Hierarchical Dirichlet Process Hidden Semi-Markov Model, mapped to a scenario-specific semantic grid, and assembled into a probabilistic DBG of behavior chains and transitions. The DBG shows that expert emergency behavior is proactive and phased: evasive action is initiated far earlier (lateral-maneuver onset at about 55 vs. 115 m in the cut-out scenario), longitudinal and lateral control are coordinated, and the vehicle is deliberately recovered after the conflict; it further localizes the bifurcation points where excellent and poor episodes diverge. Notably, self-assessed driving skill does not predict objective emergency performance, indicating that competence in safety-critical handling cannot be inferred from routine self-report. To close the loop, DBG-derived primitives are synthesized into reference trajectories that guide a model predictive controller, which reproduces expert-like earlier lateral evasion (p < 0.001) and a deliberate safety-over-compliance trade-off, and runs in a vehicle-in-the-loop deployment on a Level-3 automated test vehicle. To support reproducibility, the skill-screened expert emergency-driving dataset is publicly available at https://zenodo.org/records/20710939.
- Open Access
- 17 June 2026
- Published by Elsevier in SSRN Electronic Journal
From Human Semantics to Executable Control: Driving Behavior Graph for Automated Emergency Maneuvering
Emergency handling is among the hardest capabilities for automated driving. Safe evasion demands staged, anticipatory, and tightly coupled longitudinal-lateral actions, yet the expert logic behind such maneuvers is scarce, hard to label by skill, and locked in continuous control signals that obscure why some maneuvers succeed and others fail. This study introduces the Driving Behavior Graph (DBG), an interpretable, probabilistic representation that turns scarce expert demonstrations into reusable knowledge linking human behavioral semantics to machine-executable control. Through a two-stage procedure, we screen 25 expert drivers from 99 candidates and collect 400 driver-in-the-loop trials across four emergency scenarios (jaywalking, cut-out, cut-in, merge-in) on a virtual-reality platform. Control signals are segmented with a Hierarchical Dirichlet Process Hidden Semi-Markov Model, mapped to a scenario-specific semantic grid, and assembled into a probabilistic DBG of behavior chains and transitions. The DBG shows that expert emergency behavior is proactive and phased (earlier evasive initiation, coordinated longitudinal-lateral control, and post-conflict recovery), and it localizes the bifurcation points where excellent and poor episodes diverge; notably, self-assessed skill does not predict objective emergency performance. To close the loop, DBG-derived primitives are synthesized into reference trajectories that guide a model predictive controller, which reproduces expert-like earlier lateral evasion (p < 0.001) and a deliberate safety-over-compliance trade-off, and runs in a vehicle-in-the-loop deployment on a Level-3 automated test vehicle. To support reproducibility, the skill-screened expert emergency-driving dataset is publicly available at https://zenodo.org/records/20710939.
- 17 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessFrom Human Semantics to Executable Control: Driving Behavior Graph for Automated Emergency Maneuvering
Emergency handling is among the hardest capabilities for automated driving. Safe evasion demands staged, anticipatory, and tightly coupled longitudinal-lateral actions, yet the expert logic behind such maneuvers is scarce, hard to label by skill, and locked in continuous control signals that obscure why some maneuvers succeed and others fail. This study introduces the Driving Behavior Graph (DBG), an interpretable, probabilistic representation that turns scarce expert demonstrations into reusable knowledge linking human behavioral semantics to machine-executable control. Through a two-stage procedure, we screen 25 expert drivers from 99 candidates and collect 400 driver-in-the-loop trials across four emergency scenarios (jaywalking, cut-out, cut-in, merge-in) on a virtual-reality platform. Control signals are segmented with a Hierarchical Dirichlet Process Hidden Semi-Markov Model, mapped to a scenario-specific semantic grid, and assembled into a probabilistic DBG of behavior chains and transitions. The DBG shows that expert emergency behavior is proactive and phased (earlier evasive initiation, coordinated longitudinal-lateral control, and post-conflict recovery), and it localizes the bifurcation points where excellent and poor episodes diverge; notably, self-assessed skill does not predict objective emergency performance. To close the loop, DBG-derived primitives are synthesized into reference trajectories that guide a model predictive controller, which reproduces expert-like earlier lateral evasion (p < 0.001) and a deliberate safety-over-compliance trade-off, and runs in a vehicle-in-the-loop deployment on a Level-3 automated test vehicle. To support reproducibility, the skill-screened expert emergency-driving dataset is publicly available at https://zenodo.org/records/20710939.