I have participated and worked on different projects, others were based in Europe, others were sponsored by the NSF in the United States, while others were just part of a class that gave the opportunity to excel and specialize on a specific topic.
CESAR stands for Cost-efficient methods and processes for safety relevant embedded systems and it was the first large project that I participated, it when I first joined CSL. The goal was for the CSL to develop autonomous systems whose design would be focused on safety. Therefore, my role was to develop a nonlinear and linear dynamic model of a small scale helicopter that would lead to the completion of my diploma thesis and could later be used later in the premises of the CESAR project; the ultimate target was to have a simulation tool for the small-scale helicopter, so that off-line tests could be performed without harming the platform.
In September 2010 I started working on R3-COP as a member of CSL. The goal of the project was to develop algorithms and testbeds for robust and safe cognitive, reasoning autonomous and co-operative robotic systems at reduced cost, and the CSL in particular was responsible for providing solutions in the area of Uninhabited Underwated Vehicles (UUV). My role was twofold; primarily I had to manage the project on behalf of the CSL and assist professor K.J. Kyriakopoulos with the overall participation of the CSL in the project. Hence, I had to assure the timely delivery of each work package, as well as define the future technical and managerial requirements on behalf of the CSL. Besides the managerial work, however, I worked extensively on the programming of one of the CSL ROVs; specifically, I programmed the communication and control between a Linux-based PC-station and the robotic vehicle with the use of C++ and the Internet Communication Engine (Ice), so that advanced control schemes could be implemented.
NASCAR stands for Neuro-inspired Adaptive Sensing and Control for Agile Respone and was the first project I participated in after my arrival in the US. The goal of the project waas to develop algorithms that would bridge the gap between autonomous and semi-autonomous ground vehicles in terms of agility and speed. My role was rather restricted and my work was focused on evaluating control algorithms for satisfactory trajectory tracking by a ground vehicle. Specifically, I implemented MPC to a nonlinear vehicle dynamics model with the purpose of succesfully performing a trajectory tracking, particularly that of a double lane change. Throughout this process I became well acquainted with optimization software, and vehicle and tire dynamics.
ADAASDCAA as an acronym is my invention (for the purpose of conciseness) and stands for Advanced
Driver Assistance and Active Safety Systems through Driver’s Controllability Augmentation and Adaptation. The idea of the project was to develop control algorithms for Advanced Driver Assistance Systems (ADAS) for semi-autonomous driving that would take into consideration the driver and
increase safety of the vehicles. My main research work at Georgia Tech was performed in the premises of this project.
More specifically, I worked towards the development of an ADAS for lane-tracking by taking into consideration the “human driver dynamics”; I used the mathematical model for the two-point visual driver model and coupled them with the vehicle and steering dynamics, in order for the vehicle to perform semi-autonomous lane-tracking. The most important result of this work was that our ADAS performed significantly better than the classical ADAS that are oblivious to the driver's actions. Different optimization-based control algorithms were used in order to achieve the requested maneuver. Finally, since the driver is a parameter-varying system, i.e., diziness, reaction time, etc., adaptation techniques need to be investigated so that the developed ADAS can be satisfactorily implemented in real world vehicles.
In the Fall 2011 I was very lucky (the teams were created randomly) to be part of the “Blue Team” for the Optimal Guidance and Control class, since, first of all, I met three incredible people who later became (and still are!) good friends. The workload was distributed equally to each one of us and the result was very satisfactory. The purpose of the project was to generate the optimal trajectories for dynamic soaring (i.e., aircraft without propulsion that flies only by exploiting the existing wind fields) under static and varying wind fields. In addition to that, we had to investigate which vehicle parameters were most crucial for improving the flying capabilities of such an aircraft. The outcome of our collaborative work was more that successful, since we managed to satisfy all the requirements to the fullest, while at the time we worked on the project, the problem of optimal dynamic soaring under varying fields was not yet published. It was unfortunate that we did not submit for publication, but a copy of our project report is available upon request.
by Haddad & Chellaboina
by Horn & Johnson
by Gelfand & Fomin