Xuanpeng Zhao

The EcoCAR EV Challenge is a multi-year collegiate automotive engineering competition that brings together student teams from leading North American universities to re-engineer the 2023 Cadillac LYRIQ platform, integrating advanced battery electric propulsion systems, Connected and Automated Vehicle (CAV) technologies, and intelligent energy management solutions. Through hands-on experience with industry-standard hardware platforms including dSPACE AUTERA for real-time autonomous control development, Speedgoat systems for motor control and hardware-in-the-loop validation, and Cohda Wireless MK6 units for V2X communication, teams develop comprehensive autonomous driving capabilities while conducting rigorous testing across both controlled simulation environments and real-world driving scenarios. This collaborative initiative, supported by the U.S. Department of Energy and industry partners, aims to cultivate the next generation of automotive engineers through intensive hardware integration, closed-loop dynamometer testing, hardware-in-the-loop validation workflows, and extensive on-road testing campaigns that demonstrate the practical viability of advanced electric vehicle technologies and cooperative driving automation systems.

The escalating concern over non-tailpipe particulate matter emissions from vehicle braking systems presents a critical challenge for transportation sustainability and public health protection. This project addresses the urgent need for comprehensive real-world brake activity measurements through an innovative multi-modal sensing framework that integrates roadside LiDAR technology for high-resolution trajectory detection, supported by drone surveillance and instrumented probe vehicles for model training and validation. The framework employs a hybrid physical-machine learning approach, combining physics-based decomposition of vehicle dynamics with deep learning algorithms to isolate brake-induced deceleration from other forces such as rolling resistance, aerodynamic drag, and road gradient effects. To account for evolving vehicle technologies, the system is designed with distinct models for combustion engine vehicles and electric vehicles, each trained on powertrain-specific datasets to capture unique braking behaviors, including regenerative braking dynamics in EVs. By leveraging LiDAR-derived trajectories and material-specific emission factors calibrated against laboratory dynamometer data, the project aims to establish scalable protocols for monitoring brake wear emissions across diverse traffic conditions. The outcomes will support evidence-based policymaking by generating actionable insights into intersection-level emission hotspots, informing regulatory standards for brake materials, and facilitating the integration of adaptive traffic management strategies with emerging EV technologies to mitigate non-tailpipe pollution. Ongoing efforts focus on refining sensor fusion algorithms, expanding real-world validation across disadvantaged communities, and developing cost-effective deployment strategies for transportation agencies.

The growing demand for autonomous vehicle development platforms necessitates cost-effective, accessible solutions that enable researchers and engineers to implement and validate advanced control algorithms on production vehicles without compromising safety or requiring extensive mechanical modifications. This project aims to retrofit a KIA NIRO with comprehensive drive-by-wire capabilities, integrating electronic control systems for throttle, braking, and steering functions while maintaining the vehicle's original safety architecture and enabling seamless integration with autonomous driving software platforms. Through the implementation of electronic control units, sensor interfaces, and actuator systems, the conversion will establish a reliable foundation for autonomous vehicle research, leveraging the NIRO's electric vehicle architecture to create a fully controllable platform suitable for algorithm development, real-world testing, and validation of connected and automated vehicle technologies. The project will utilize either custom STM32-based signal interception systems or commercial solutions like PolySync DriveKit to ensure compatibility with industry-standard development tools while maintaining the flexibility to integrate specialized hardware and software components for autonomous vehicle research applications. By converting the KIA NIRO to full drive-by-wire operation, this initiative will provide a robust, cost-effective autonomous vehicle development platform that supports comprehensive testing of perception, planning, and control algorithms while enabling safe human oversight and emergency intervention capabilities.

The escalating concern over freight movement efficiency and environmental impacts in urban environments necessitates innovative solutions that can optimize traffic signal operations while considering the unique operational characteristics of commercial vehicles. This infrastructure-based Freight Signal Priority project addresses the critical need for comprehensive real-world freight movement optimization through an advanced roadside sensing framework that integrates LiDAR technology for high-resolution vehicle detection, classification, and state estimation without requiring any onboard equipment modifications to freight vehicles. The framework will employ sophisticated point cloud processing algorithms and machine learning techniques to distinguish freight vehicles from other traffic participants, predict arrival times with high accuracy, and generate priority requests that consider factors such as vehicle weight, emission profiles, and delivery urgency while maintaining balance with general traffic flow. Through the implementation of strategically positioned roadside LiDAR sensors and standardized communication protocols including NTCIP 1202 and NTCIP 1211, the system will integrate seamlessly with existing traffic signal controllers to provide real-time signal optimization that balances freight priority with overall urban mobility efficiency. The project aims to demonstrate robust performance against various traffic scenarios while maintaining low computational overhead suitable for real-time deployment, ultimately establishing scalable protocols for monitoring and prioritizing freight movement that can be readily adopted by transportation agencies seeking to improve urban freight efficiency while maintaining equitable traffic management for all road users.

The increasing deployment of autonomous vehicles in critical applications introduces new security vulnerabilities, particularly the risk of sensor manipulation attacks that can compromise vehicle safety and operational integrity by exploiting physical interfaces to GPS, IMU, and other critical sensors. This threat poses significant challenges to the widespread adoption of autonomous systems, as traditional cybersecurity approaches prove inadequate against physical-world attacks that can cause vehicles to deviate from intended trajectories or even crash. AVMON addresses these security concerns through a comprehensive framework that combines optimized Extended Kalman Filter parameter tuning specific to vehicle dynamics and operational environments, residual learning using neural networks to compensate for nonlinear effects and unmodeled disturbances, and advanced change detection algorithms to reduce false positives while maintaining real-time performance. The system demonstrates substantial improvements in attack detection accuracy, achieving detection rates of 98-100% compared to existing defenses while maintaining low computational overhead suitable for deployment on resource-constrained embedded platforms. Through extensive validation using both simulation environments and hardware testbeds for ground and aerial vehicles, AVMON proves its effectiveness against sophisticated attack scenarios including GPS spoofing, IMU manipulation, and adaptive attacks designed to evade detection, with the potential for technology transfer to enhance security across diverse autonomous vehicle applications.

Developed a traffic monitoring system using LiDAR...

Driver Intention Prediction for Cooperative Highway Ramp Merging...

Game-Theoretic Control Based Highway Ramp Merging...

Vehicle-to-Cloud ADAS for Cooperative Highway Ramp Merging...

Integrated Simulation Platform for Quantifying the Traffic-Induced Environmental and Health Impacts...

Co-simulation platform for modeling and evaluating connected and automated vehicles and human behavior in mixed traffic...

Evaluating cybersecurity risks of cooperative ramp merging in mixed traffic environments...