Behavior planning for self driving cars.

- Implemented full stack vehicle autonomy software, including behavior planning utilizing scalable behavior tree framework, highly-efficient semi-analytic trajectory and motion planning, high-speed localization, traffic state estimation, multi-vehicle coordination and deconfliction, LiDAR perception (PCL), and safety protection modules.- Build, grow, train, mentor, manage, and lead a highly efficient autonomy team to iterate and mature vehicle autonomy designs and software through consistent release cycles and software development process (version control, code review, etc.). - Key contributor in overall Hyperloop command and control (C&C) system design, including C&C architecture, operation concept and strategy, vehicle and control center coordination under contingencies, and safety analysis. - Coordinate and collaborate across teams to navigate and identify the challenges/risks, perform feasibility and trade studies, down select long-term technology roadmap, define short-term plans, break down actionable tasks, and execute efficiently.

- Developed the Hyperloop pod localization system, including onboard and wayside sensors network designs, onboard state estimation algorithms, and software-in-loop (SIL) evaluation. - Developed the multi-pod convoying system enabling high throughput, including convoying motion control algorithms, stability analysis, system architecture design, sensor and communication specifications, and SIL evaluation. - Developed the safety critical collision avoidance system, including state machine for decision making, onboard and wayside system architecture, communication and sensor specifications, SIL testing, and safety analysis. - Developed the adaptive controller for Hyperloop active magnetic levitation system, including system dynamic modeling and identification, signal processing and filtering, state estimation, control law development, and hardware-in-loop testing.

My Ph.D research was to develop and promote novel technologies to revolutionize traditional transportation. In the process, I developed skill sets in traffic modeling, state estimation, control and optimization, machine learning, data mining, and smart traffic sensors. Vehicle Detection using Passive Infrared Sensors and Unsupervised Learning - Developed a low-cost and energy-efficient vehicle detection system combining low-resolution passive infrared cameras (4 x 32 pixels) and ultrasonic sensors. - Developed unsupervised machine learning algorithms (utilizing DBSCAN, Hough transform, regression, low-dimensional segmentation) for accurate vehicle detection and speed estimation. - Validated the high performance of the developed vehicle detection system in field experiments against computer vision (OpenCV) results on video data.Modeling and Estimation of Mixed Human-piloted and Autonomous Vehicles - Developed a two-class macroscopic traffic model for simulating the mixed behavior of human-piloted and autonomous vehicles. - Developed multi-model particle filtering algorithms for estimating the mixed traffic state using fixed-position distributed traffic flow data. - Validated the performance of the traffic estimation algorithm in a microsimulation environment with simulated human-piloted and autonomous vehicles.Modeling, Estimation, and Control of Nonlinear Traffic Flows on Networks - Developed nonlinear dynamic models of traffic flows on transportation networks with well-behaved junction models using partial differential equations. - Developed an efficient convex optimization based numerical framework for real-time traffic estimation and optimal control networks. - Developed various types of state estimation algorithms (ensemble Kalman filtering, particle filtering) for estimating traffic velocity field, traffic density, queue length, and travel time.

Machine learning and data mining in mobility applications. Trip Prediction for Divvy Bike Sharing Program (temporal-spatial data mining)- Visualized and analyzed over 750,000 Divvy trips in Chicago collected over six months. - Developed different distance measures in clustering algorithms to address the data sparsity and high-dimensionality. - Implemented linear and ARX regression models for prediction of bike trips. Differential Privacy in Traffic Flow data- Analyzed privacy risks of public traffic volume data set (over 20 million entries) collected from over 1000 traffic detectors in California PeMS system. - Developed differential privacy masking mechanism to apply to the PeMS traffic dataset before publishing to preserve traveller privacy in the public PeMS traffic data. - Analyzed and quantified the trade off between the data quality and privacy protection levels.

- Focused on mathematical proof of the generalization of nonlinear traffic flow models from single link to a network with well-behaved junction models. - Developed MATLAB toolbox based on the proposed modeling framework, and ran simulations to evaluate performance of proposed framework in traffic control applications. - Initial vehicle detection algorithm development from low resolution data collected by the developed second generation smart passive infrared traffic sensors.

Research on modeling, robust and optimal control of traffic flow on networks in Distributed Sensing Systems lab at KAUST. - Developed an optimal control framework for traffic flow on networks using linear programming, quadratic programing, and mixed integer programming. - Developed a robust traffic controller on networks by incorporating boundary traffic flow data uncertainty. - Developed and released a MATLAB toolbox for optimal and robust traffic flow control on networks. Published two journal articles and two conference proceedings.

Led a team to build and test an autonomous fixed-wing UAV in the Distributed Sensing Systems lab at KAUST.- Built an autonomous fixed-wing aircraft, by integrating the open source Ardupilot embedded computing unit, IMU, GPS, pitot tube, ultrasonic sensor, barometer, XBee telemetry module, battery, and rotors.- Embedded programming (on sensor interfacing, sensor fusion, controller implementation, and serial communication) on Ardupilot Mega board and Gumstix microcomputer (embedded linux system).- Conducted autopilot controller parameter tuning, field flight tests, and flight data analysis.