Qi Chen
University of California-Irvine, CA·Computer Science·Early-career
Self-driving cars, delivery drones and robots rely on an artificial intelligence (AI) software stack as their brain: it reads sensors, decides what is in the world and chooses how to move. If that brain can be fooled, the consequences are physical rather than merely digital.
- Fields
Computer Science
Engineering
- Topics
autonomous driving
connected and autonomous vehicles
cyber-physical systems
- Grants
- 2 known grants, through Nov 2027
What they work on
- Develop system-to-AI and AI-to-system mapping methods that generate physical-layer attacks on autonomous vehicle AI while keeping them physically realizable and semantically equivalent across diverse real system designs.
- Design system-level defenses based on new classes of physical invariants, relationships that must hold when sensors report honestly, and on attack-resilient sensor fusion that exploits whole-system properties.
- Build defenses that draw on the surrounding operating ecosystem, including other vehicles and roadside infrastructure support, to detect or absorb attacks on a single autonomous vehicle.
- Create a simulation-based evaluation platform with uniform, extensible support for developing and comparing attacks and defenses on autonomous cyber-physical systems, for use in research and coursework.
- Identify new attack surfaces introduced when connected vehicles and roadside units share perception data in collaborative autonomous driving, combining vehicle-to-everything security analysis, adversarial attacks on AI perception, and formal verification.
- Propose and evaluate mitigation and defense mechanisms for collaborative autonomous driving, using robust control and end-to-end system evaluation together with partner groups in the United States and India.
More about the research
Most prior security research has probed the AI algorithms on their own, in isolation from the vehicle, the sensors and the road around them. The work here instead treats these machines as whole systems, asking what an attacker can actually achieve from the physical world and how a complete system can be designed to survive such attempts.
Why it matters
As autonomous vehicles and delivery robots move into everyday traffic, flaws in their AI software can translate directly into crashes. Analyzing attacks that originate in the physical world and designing system-level defenses, including protections for vehicles that share sensor data with each other and with roadside units, could make these systems harder to mislead. The projects also produce public datasets, simulation platforms and teaching materials for autonomous systems security.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Justin Ruths
University of Texas at Dallas, TX·Engineering·Early-career
Represent each autonomy-stack module (mapping, perception, planning, control) and the vehicle's physical dynamics as hybrid zonotopes, sets made of unions of convex polytopes, to create one common form.
Sayan Mitra
University of Illinois Urbana-Champaign, IL·Computer Science
Develop refinement algorithms that synthesize perception contracts, machine-checkable bounds on how wrong an AI perception module can be, together with the operating design domains in which they hold, for pipelines that fuse camera and inertial sensor data.
Zhi-Li Zhang
University of Minnesota, MN·Computer Science
Develop an O-RAN-based framework in which 5G networks use vehicle trajectory and speed to proactively cut delays from poor radio signals and base-station handovers during remote driving.
The school
University of California-Irvine: 46 labs on PI Pages, $645M a year on research, 524 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.