Cyber-Physical Systems (CPS) are typically composed of interconnected hardware and software components, which individually may not be inherently highly reliable or secure. However, several CPS applications demand a high degree of safety, security, and reliability. Thus, the fundamental problem is constructing highly dependable CPS applications from building blocks that are, in themselves, not inherently reliable.
Chorus will develop rigorous, scientific mechanisms to enable CPS resilience against a large universe of perturbations. Our application domain is Connected and Autonomous Transportation Systems (CATS) and thus, the benefits of CHORUS will be demonstrated through improvements in safety and security in this domain.
We will achieve goals of CHORUS through three interacting intellectually challenging thrusts in the project.
In terms of broader impact, the greatest impact will be that CPS owners will gain a higher degree of trust in the operation of the CPS and policy-makers will understand what level of cooperation among multiple stakeholders in a CPS to incentivize. We will create compelling demonstrations of CHORUS on a connected vehicle testbed distributed between our academic institutions and our industrial partner GM. We will also organize an annual student security competition and develop two MOOCS, both having foundational material on resilient CPS and one focusing more on the CATS application domain.
March 2025
The paper addresses the problem of adapting pre-trained video large lanuage models to downstream tasks involving additional input modality or different data type. This work falls under Thrust 3 of the CHORUS center.
Citation: Zhuoming Liu, Yiquan Li, Khoi Duc Nguyen, Yiwu Zhong, and Yin Li. PAVE: Patching and Adapting Video Large Language Models. IEEE/CVF Computer Vision and Pattern Recognition Conference (CVPR) 2025. (acceptance rate: 22.1%)
Software Release: We have open-sourced the code for our CVPR 2025 paper at https://github.com/dragonlzm/PAVE
March 2025
Our collaborative paper between Purdue and partner institutions has been accepted to MobiSys 2025. This work presents an adaptive 3D object detection system that dynamically reconfigures execution based on runtime contention and content characteristics. This research falls under Thrust 3 of the CHORUS center.
Citation: Pengcheng Wang, Zhuoming Liu, Shayok Bagchi, Ran Xu, Saurabh Bagchi, Yin Li, and Somali Chaterji. Agile3D: Adaptive Contention- and Content-Aware 3D Object Detection for Embedded GPUs. ACM International Conference on Mobile Systems, Applications, and Services (MobiSys) 2025. (acceptance rate: 18.0%)
November 19-21, 2024
Event Link: Grand Challenges Workshop 2024
Location: Wright Center, Martell Forest (Reception on the 19th); Recreational Sports Center (20th and 21st); Shively Club (Banquet on the 20th), Purdue University, West Lafayette, IN
Kickstarting our NSF-funded CHORUS Center, this workshop will bring together leaders in resilient cyber-infrastructures, cyber-physical systems, and socio-technical resilience. Industry and academic speakers, along with federal program managers, will introduce the vision and goals through foundational techniques and real-world case studies to strengthen adaptive and resilient critical infrastructure.