Juntong Peng
Research Subjects
Juntong Peng has a total of 32 co-authors. Here are the top co-authors by number of shared publications:
- Can Cui
Civil Engineering, Purdue University, USA
3 shared publications - Ziran Wang
Purdue University, West Lafayette, IN, USA
3 shared publications - Yupeng Zhou
Purdue University System
2 shared publications - Zichong Yang
Purdue University
2 shared publications - Abin Mathew
Purdue University
1 shared publication - Agnideven Palanisamy Sundar
Department of Computer Science, Purdue University-Indianapolis, Indianapolis, IN, USA
1 shared publication
Latest Publications
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (22) , 53179-53190
VLM-DC: Vision--Language Model for Automotive Data Collection
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (22) , 53179-53190
VLM-DC: Vision--Language Model for Automotive Data Collection
- 17 September 2026
- Published by Springer Nature
- p. 540-556
QuantV2X: A Fully Quantized Multi-agent System for Cooperative Perception
- 17 September 2026
- Published by Springer Nature
- p. 540-556
QuantV2X: A Fully Quantized Multi-agent System for Cooperative Perception
- Open Access
- 21 April 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Internet of Things
- Vol. 7 (3) , 1-28
A Hierarchical Test Platform for Vision Language Model (VLM)-Integrated Real-World Autonomous Driving
Vision-Language Models (VLMs) have demonstrated significant promise for autonomous driving due to their powerful multimodal reasoning capabilities. However, adapting VLMs from generic data to safety-critical driving contexts introduces a notable challenge known as domain shift. Existing simulation-based and dataset-driven evaluation approaches struggle to accurately replicate real-world complexities, lacking repeatable closed-loop evaluation and flexible scenario manipulation. Furthermore, current real-world testing platforms typically focus on isolated modules and do not support comprehensive interaction with VLM-based systems. Consequently, there is a critical need for a holistic testing architecture capable of integrating perception, planning, and control modules, accommodating VLM-based systems, and supporting configurable real-world testing scenarios. In this article, we address this critical gap by proposing a hierarchical real-world test platform specialized in the rigorous evaluation of VLM-integrated autonomous driving systems. Specifically, our platform features have: a lightweight, structured, and low-latency middleware pipeline specialized for seamless VLM integration; a hierarchical modular architecture enabling flexible substitution between conventional and VLM-based autonomy components, providing exceptional deployment flexibility for rapid experimentation; and sophisticated closed-loop scenario-based testing capabilities on a controlled test track, facilitating comprehensive evaluation of the entire full-stack VLM-integrated autonomous driving pipeline, from perception, reasoning, decision-making, and planning to final vehicle maneuvers. Through an extensive real-world case study, we demonstrate the effectiveness of our platform in evaluating the performance and robustness of VLM-integrated autonomous driving under diverse realistic conditions. Project page and codes: https://github.com/YupengZhouPurdue/VLMTest .
- 21 April 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Internet of Things
- Vol. 7 (3) , 1-28
Open AccessA Hierarchical Test Platform for Vision Language Model (VLM)-Integrated Real-World Autonomous Driving
Vision-Language Models (VLMs) have demonstrated significant promise for autonomous driving due to their powerful multimodal reasoning capabilities. However, adapting VLMs from generic data to safety-critical driving contexts introduces a notable challenge known as domain shift. Existing simulation-based and dataset-driven evaluation approaches struggle to accurately replicate real-world complexities, lacking repeatable closed-loop evaluation and flexible scenario manipulation. Furthermore, current real-world testing platforms typically focus on isolated modules and do not support comprehensive interaction with VLM-based systems. Consequently, there is a critical need for a holistic testing architecture capable of integrating perception, planning, and control modules, accommodating VLM-based systems, and supporting configurable real-world testing scenarios. In this article, we address this critical gap by proposing a hierarchical real-world test platform specialized in the rigorous evaluation of VLM-integrated autonomous driving systems. Specifically, our platform features have: a lightweight, structured, and low-latency middleware pipeline specialized for seamless VLM integration; a hierarchical modular architecture enabling flexible substitution between conventional and VLM-based autonomy components, providing exceptional deployment flexibility for rapid experimentation; and sophisticated closed-loop scenario-based testing capabilities on a controlled test track, facilitating comprehensive evaluation of the entire full-stack VLM-integrated autonomous driving pipeline, from perception, reasoning, decision-making, and planning to final vehicle maneuvers. Through an extensive real-world case study, we demonstrate the effectiveness of our platform in evaluating the performance and robustness of VLM-integrated autonomous driving under diverse realistic conditions. Project page and codes: https://github.com/YupengZhouPurdue/VLMTest .
- 1 January 2026
- Published by Purdue University Press
RampCast Phase II: Connected Vehicles Traffic Management Application on Indiana Highways
This project advances connected vehicle applications by developing and testing an enhanced RampCast system, a comprehensive traffic management system using C-V2X technology for Indiana highways. The system features a dual-mode architecture integrating both short-range (PC5) and long-range cellular (Uu) Cellular Vehicle-to-Everything (C-V2X) communication pathways, utilizing commercial-grade Cohda MK6 hardware and adhering to SAE J2735 standards to ensure interoperability. A key innovation is the integration of an AI-based prioritization framework, which leverages a large language model to enhance the contextual relevance of traffic messages. This AI system introduces two intelligent agents: one to dynamically estimate the appropriate display distance for an event based on its severity, and another to prioritize the order of messages based on urgency and potential driver impact. Field tests conducted on I-65 and I-70 in Indianapolis validated the system’s hybrid design. Results confirmed that the PC5 link provides very low latency (around 25 ms), ideal for time-critical alerts, while the Uu link ensures highly reliable coverage in complex environments, albeit with higher latency (around 45 ms). The AI framework was successfully shown to reorder and present messages based on real-time context, improving the clarity and usefulness of information provided to the driver. These findings support a hybrid C-V2X architecture as a robust model for future smart highway deployments.
- 1 January 2026
- Published by Purdue University Press
RampCast Phase II: Connected Vehicles Traffic Management Application on Indiana Highways
This project advances connected vehicle applications by developing and testing an enhanced RampCast system, a comprehensive traffic management system using C-V2X technology for Indiana highways. The system features a dual-mode architecture integrating both short-range (PC5) and long-range cellular (Uu) Cellular Vehicle-to-Everything (C-V2X) communication pathways, utilizing commercial-grade Cohda MK6 hardware and adhering to SAE J2735 standards to ensure interoperability. A key innovation is the integration of an AI-based prioritization framework, which leverages a large language model to enhance the contextual relevance of traffic messages. This AI system introduces two intelligent agents: one to dynamically estimate the appropriate display distance for an event based on its severity, and another to prioritize the order of messages based on urgency and potential driver impact. Field tests conducted on I-65 and I-70 in Indianapolis validated the system’s hybrid design. Results confirmed that the PC5 link provides very low latency (around 25 ms), ideal for time-critical alerts, while the Uu link ensures highly reliable coverage in complex environments, albeit with higher latency (around 45 ms). The AI framework was successfully shown to reorder and present messages based on real-time context, improving the clarity and usefulness of information provided to the driver. These findings support a hybrid C-V2X architecture as a robust model for future smart highway deployments.
- 1 January 2026
- Published by Purdue University Press
CAV Pilot Development and Deployment in Midwest Winter
This project addresses the critical challenges of deploying Connected and Automated Vehicles (CAVs) in adverse Midwest winter conditions by developing a novel hierarchical framework that integrates Large Vision-Language Models for robust, context-aware decision-making. Real-world pilot experiments demonstrate that this architecture significantly enhances vehicle safety and personalization, effectively translating abstract human commands into precise control adjustments on low-friction, snow-covered surfaces. Furthermore, a specialized high-fidelity dataset capturing unique environmental edge cases was established to bridge the domain gap in existing resources and support the future development of resilient all-weather autonomous technologies.
- 1 January 2026
- Published by Purdue University Press
CAV Pilot Development and Deployment in Midwest Winter
This project addresses the critical challenges of deploying Connected and Automated Vehicles (CAVs) in adverse Midwest winter conditions by developing a novel hierarchical framework that integrates Large Vision-Language Models for robust, context-aware decision-making. Real-world pilot experiments demonstrate that this architecture significantly enhances vehicle safety and personalization, effectively translating abstract human commands into precise control adjustments on low-friction, snow-covered surfaces. Furthermore, a specialized high-fidelity dataset capturing unique environmental edge cases was established to bridge the domain gap in existing resources and support the future development of resilient all-weather autonomous technologies.
- 4 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 9 (11) , 10676-10683
Graph-Based Decentralized Task Allocation for Multi-Robot Target Localization
We introduce a new graph neural operator-based approach for task allocation in a system of heterogeneous robots composed of Unmanned Ground Vehicles (UGVs) and Unmanned Aerial Vehicles (UAVs). The proposed model, GATAR , or G raph A ttention T ask A llocato R aggregates information from neighbors in the multi-robot system, with the aim of achieving globally optimal target localization. Being decentralized, our method is highly robust and adaptable to situations where the number of robots and the number of tasks may change over time. We also propose a heterogeneity-aware preprocessing technique to model the heterogeneity of the system. The experimental results demonstrate the effectiveness and scalability of the proposed approach in a range of simulated scenarios generated by varying the number of UGVs and UAVs and the number and location of the targets. We show that a single model can handle a heterogeneous robot team with a number of robots ranging between 2 and 12 while outperforming the baseline architectures.
- 4 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 9 (11) , 10676-10683
Graph-Based Decentralized Task Allocation for Multi-Robot Target Localization
We introduce a new graph neural operator-based approach for task allocation in a system of heterogeneous robots composed of Unmanned Ground Vehicles (UGVs) and Unmanned Aerial Vehicles (UAVs). The proposed model, GATAR , or G raph A ttention T ask A llocato R aggregates information from neighbors in the multi-robot system, with the aim of achieving globally optimal target localization. Being decentralized, our method is highly robust and adaptable to situations where the number of robots and the number of tasks may change over time. We also propose a heterogeneity-aware preprocessing technique to model the heterogeneity of the system. The experimental results demonstrate the effectiveness and scalability of the proposed approach in a range of simulated scenarios generated by varying the number of UGVs and UAVs and the number and location of the targets. We show that a single model can handle a heterogeneous robot team with a number of robots ranging between 2 and 12 while outperforming the baseline architectures.