One of the critical components of pavement structures are shoulders which provide lateral support to the entire pavement structure. The vertical elevation between pavement structure surface and adjacent shoulder surface results in roadway edge drop-off. Shoulder drop-off can affect the vehicle’s stability and driver’s ability to handle the vehicle resulting in fatalities or damage, which is also one of transportation agencies’ most cited accident-related highway conditions. The maintenance staff of state transportation departments visually inspect the shoulders at near highway speeds, which might be challenging and involves human judgment, affecting the assessment’s accuracy and reliability. In this work, we provide a lost-cost solution to automate the process of shoulder drop-off assessment. Our approach analyzes the point cloud data acquired from a 3D solid-state LIDAR sensor mounted on a car traveling at highway speeds. We have explored statistical, machine learning, and end-to-end deep learning methods to predict the vertical elevation of shoulder drop-off with accuracy greater than 95$\%$. We have validated the accuracy of proposed learning-based methods with actual data collected using inexpensive LIVOX LiDAR and an Astra camera along various highway sections.

The security of cyber-physical systems, such as vehicle platoons, is critical to ensuring their proper operation and acceptance to society. In platooning, vehicles follow one another according to an agreed upon control law that determines vehicle separation. It has been shown that a vehicle within a platoon and under the control of a malicious actor could cause collisions involving, or decrease the efficiency of, surrounding vehicles. In this paper we focus on detecting, identifying and mitigating so called destabilizing attacks that could cause vehicle collisions. Our approach is decentralized and requires only local sensor information for each vehicle to identify the vehicle responsible for the attack and then deploy an appropriate mitigating controller that prevents collisions. A Deep Learning approach (Convolutional Neural Network) with various data preprocessing techniques are used to detect and identify the malicious vehicle. Results indicate that with noise upto 30% in range/relative speed data we achieve an accuracy upto 96.3%. Also, once the adversarial vehicle is localized, we derive conditions for controller gains using Routh Hurwitz criterion to mitigate the attack and ensure stability of the platoon. Realistic simulator CARLA and MATLAB simulation results validate the effectiveness of our proposed approaches