Ao Li

Ao Li

Auto-generated profile, based on publicly available information

Ao Li

Washington University in St. Louis, St. Louis, USA

Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
24 December 2021 in IEEE Transactions on Intelligent Transportation Systems
Verified
Journal Article
Closed access
ALAo Li
Ao Li
SCShitao Chen
Shitao Chen
LSLiting Sun
Liting Sun
NZNanning Zheng
Nanning Zheng
MTMasayoshi Tomizuka
Masayoshi Tomizuka
WZWei Zhan
Wei Zhan
The core value of simulation-based autonomy tests is to create densely extreme traffic scenarios to test the performance and robustness of the algorithms and systems. Test scenarios are usually designed or extracted manually from the real-world data, which is inefficient with a remarkable domain gap compared with testing in real scenarios. Therefore, it is crucial to automatically generate realistic and diverse dynamic traffic scenarios making autonomy tests efficient. Moreover, scenario generation is expected to be interpretable, controllable, and diversified, which can be hard to achieve simultaneously by methods based on rules or deep networks. In this paper, we propose a dynamic traffic scenario generation method called SceGene, inspired by genetic inheritance and mutation processes in biological intelligence. SceGene applies biological processes, such as crossover and mutation, to exchange and mutate the content of scenarios, and involves the natural selection process to control generation direction. SceGene has three main parts: 1) a new representation method for describing the traffic scenarios' feature; 2) a new scenario generation algorithm based on crossover, mutation, and selection; and 3) an abnormal scenario information repair method based on the microscopic driving model. Evaluation on the public traffic scenario dataset shows that SceGene can ensure highly realistic and diversified scenario generation in an interpretable and controllable way, significantly improving the efficiency of the simulation-based autonomy tests.

29 May 2023
Verified
Conference Paper
Open access
SSShanghao Shi
Shanghao Shi
YXYang Xiao
Yang Xiao
CDChanglai Du
Changlai Du
HSHasan Shahriar
Hasan Shahriar
ALAo Li
Ao Li
WLWenjing Lou
Wenjing Lou

2 December 2024
Verified
Conference Paper
Open access
ZYZhiyuan Yu
Zhiyuan Yu
ALAo Li
Ao Li
RWRuoyao Wen
Ruoyao Wen
YCYijia Chen
Yijia Chen
NZNing Zhang
Ning Zhang
Autonomous vehicles (AVs) empowered by deep neural networks (DNNs) are bringing transformative changes to our society. However, they are generally susceptible to adversarial attacks, especially physically realizable perturbations that can mislead perception and cause catastrophic outcomes. While existing defenses have shown success, there remains a pressing need for improved robustness while maintaining efficiency to meet real-time system operations. To tackle these challenges, we introduce PhySense, a complementary solution that leverages multi-faceted reasoning for misclassification detection and correction. This defense is built on physical characteristics, including static and dynamic object attributes and their interrelations. To effectively integrate these diverse sources, we develop a system based on the conditional random field that models objects and relationships as a spatial-temporal graph for holistic reasoning on the perceived scene. To ensure the defense does not violate the timing requirement of the real-time cyber-physical control loop, we profile the run-time characteristics of the workloads to parallelize and pipeline the execution of the defense implementation. The efficacy of PhySense is experimentally validated through simulations of datasets and real-world driving tests. It also demonstrates resiliency against adaptive attacks, and the potential of applying underlying principles to other modalities beyond vision.

26 November 2023
Verified
Conference Paper
Open access
ALAo Li
Ao Li
SCSinyin Chang
Sinyin Chang
GLGuorui Li
Guorui Li
YCYuanhaur Chang
Yuanhaur Chang
NFNathan Fisher
Nathan Fisher
TCThidapat (Tam) Chantem
Thidapat (Tam) Chantem
Inspired by natural swarms, swarm robotics systems are used in safety-critical tasks due to their scalability, redundancy, and adaptability. However, their design exposes them to two primary vulnerabilities. First, their homogeneity makes them vulnerable to large-scale attacks. Second, logical flaws within swarm algorithms can be exploited, leading to mission failures or crashes. While existing studies can effectively identify these vulnerabilities using system testing and verification, they are often time-consuming and might require repetition following software updates. To this end, we propose a complementary, two-level diversification approach. The first level tackles system homogeneity through software diversification. The second level introduces algorithmic randomness to minimize the exploitability of logical flaws. By leveraging a social force model, we can ensure that the introduced randomized behaviors do not compromise safety. Our evaluations show that the performance overheads remain within acceptable limits, notably at 2% for missions characterized by self-organizing behaviors.

1 March 2024 in Real-Time Systems
Verified
Journal Article
Closed access
PDF file
MSMarion Sudvarg
Marion Sudvarg
ZSZhuoran Sun
Zhuoran Sun
ALAo Li
Ao Li
CGChris Gill
Chris Gill
NZNing Zhang
Ning Zhang

11 November 2024
Verified
Conference Paper
Open access
SCSinyin Chang
Sinyin Chang
ALAo Li
Ao Li
EJEvin Jaff
Evin Jaff
YCYuanhaur Chang
Yuanhaur Chang
JWJinwen Wang
Jinwen Wang
HHHsu-Chun Hsiao
Hsu-Chun Hsiao
The current state of healthcare is challenged by the widespread use of legacy systems, which often lack the advanced security features necessary to combat modern cyber threats. However, our investigation reveals that hospitals are often reluctant to upgrade their systems, which is primarily due to the following factors: low tolerance to service downtime and patch errors, platform diversities, and the resource constraints of embedded medical devices. Motivated by these challenges, we propose AdapSan, a framework that allows users to dynamically deploy input sanitization schemes for kernel according to the risk profiles of the target system. AdapSan utilizes eBPF as the key technique to enable dynamic reconfiguration of defense strategies, thereby avoiding downtime. Additionally, it uses eBPF's verification mechanisms to sandbox patch code, minimizing the impact of potentially buggy patches. Furthermore, the platform-agnostic nature of eBPF bytecode and its Just-In-Time (JIT) compilation facilitate the deployment of this solution across various platforms while maintaining the execution speed of native code. AdapSan incorporates two types of insertion schemes, offering a balance between usability and code complexity. Preliminary evaluations of AdapSan indicate that the patching time can be less than 19 milliseconds and the average runtime overhead post-patching can be maintained below 36%.