Research Projects | Efe Yarbasi, PhD

Influence of Vehicle Active Safety Technologies on Roadway Departures (NCHRP 22-68)

Wed, 01 Jul 2026 00:00:00 +0000

Sponsor: National Cooperative Highway Research Program (NCHRP 22-68) · Prime: Texas A&M Transportation Institute (TTI) · Period: awarded 2026 (36-month subcontract)

Evaluating the Influence of Vehicle Active Safety Technologies on Roadway Departures.

Roadway departures are among the most severe crash types. The project asks which roadway departure crashes today’s vehicle active safety technologies are, and are not, preventing, to inform both vehicle and roadway countermeasures.

My contributions

UMTRI lead on the subcontract.
Co-lead the data analysis plan and its execution.
The analysis centers on “crashes not prevented”, combining GM Safety Content, naturalistic driving data, and national crash databases.

Predictive Aerial Coverage Validation

Wed, 01 Jul 2026 00:00:00 +0000

Program: Mcity / , University of Michigan, with industry partners · Period: 2026–2027 (ongoing)

Cellular networks are increasingly used to support low-altitude drone operations, for command and control and for situational awareness. Coverage predictions for the airspace above the ground are hard to trust without independent measurement. This project independently validates a predictive lower-airspace 4G/5G coverage model.

My role

Technical lead for the independent validation.
Quantify prediction accuracy and its uncertainty, and characterize the conditions under which prediction error grows, such as altitude, terrain, and frequency band.
Measurement flights follow the published ACJA Reference Method, and hold-out flights are kept separate from any model tuning.

Results will be shared through peer-reviewed publication, subject to sponsor review.

Trip Recorder Analysis (TRA)

Wed, 01 Jan 2025 00:00:00 +0000

Sponsor: National Highway Traffic Safety Administration (NHTSA) · Period: 2025 to present

Naturalistic trip recorders collect far more driving data than people can review by hand. This project develops methods to automate the analysis of trip recorder data so that it can support safety performance assessment of automated driving systems (ADS) and driver assistance (ADAS).

My contributions

Lead development of an open-source scenario labeling and retrieval toolset for automatically identifying and retrieving driving scenarios from multi-system trip recorder datasets.
Lead the work on data standardization and common data formats, so that recordings from different systems can be processed together.
Lead the scenario clustering methods, and plan the tool for open-source distribution and adoption by third parties.

A trip divided into driving scenarios (top) and the shorter potential-conflict scenarios within them (bottom). Figure from the project’s knowledge synthesis report to NHTSA.

Assessing Driving Performance Using Refined Ground-Truth Trip Recorders

Wed, 03 Jul 2024 00:00:00 +0000

Sponsor: National Highway Traffic Safety Administration (NHTSA) · Period: through 2024

The project used instrumented vehicles as “ground-truth” trip recorders to assess driving performance, especially around intersections, and refined how accurately such recorders capture the surrounding traffic.

My contributions

Worked on radar tracking and sensor fusion, combining radar, camera, and LiDAR outputs to improve accuracy in intersection-based safety assessments.
Co-author of the project’s final report to NHTSA (LeBlanc et al., 2024; in review).

Intersection Right-of-Way ADAS Technology Assessment

Tue, 02 Jul 2024 00:00:00 +0000

Sponsor: National Highway Traffic Safety Administration (NHTSA) · Period: 2024 to 2027

Intersection right-of-way crashes, such as left turns across path and crossing-path collisions, are a growing focus for driver assistance technology. This project assesses intersection right-of-way ADAS, including the risks and unintended consequences such systems could introduce.

My contributions

Led the Hazard Analysis and Risk Assessment (HARA), applying ISO 26262 functional safety, DFMEA, STPA, and SOTIF methods to identify, analyze, and prioritize risks.
Organized structured engagement with outside experts from the automotive industry, insurance, law enforcement, crash data and roadway operations, and human factors.

The vehicle’s position at the moment the light turns yellow determines which ADAS decision rules apply. Figure from the project report to NHTSA.

High-Speed AEB Crash Scenarios and Simulation-Based Safety Benefit Assessment

Mon, 01 Jul 2024 00:00:00 +0000

Sponsor: Toyota Motor North America · Period: 2024 to 2026 (completed)

Full title: High Speed AEB Crash Scenarios and Simulation-Based Assessment Including Driver Interactions and Safety Benefit Estimation.

Automatic Emergency Braking (AEB) is increasingly expected to work at higher speeds, where crash configurations and driver responses differ from the low-speed cases most systems were first designed for. The project estimated how much real-world benefit higher-speed AEB could deliver.

My contributions

Led development of a MATLAB-based simulation toolchain for assessing AEB safety benefits at higher speeds.
Built the crash scenario library and ran counterfactual simulations of real-world crashes from NHTSA’s Crash Investigation Sampling System (CISS), with and without AEB intervention.
Integrated injury risk modeling and produced the safety benefit estimates.

Text Mining of Publications

Tue, 01 Jan 2019 00:00:00 +0000

In this project at the Aerospace Systems Design Laboratory under the supervision of Prof. Dimitri Mavris and Dr. Olivia Pinon Fischer, I employed sophisticated topic modeling and text mining techniques to dissect and categorize the lab’s extensive portfolio of academic publications. By meticulously processing and analyzing a vast dataset, I was able to identify and characterize the various research collectives and their thematic clusters within the lab. Leveraging advanced algorithms and machine learning tools, I extracted pivotal patterns and discerned trends that illuminated the lab’s research trajectory and thematic development. This analytical deep dive not only showcased my proficiency in data analysis with Python and R, but also emphasized my capabilities in data visualization and the intricate fields of text mining and topic modeling. Derived insights were presented to lab administration.

NATO STO AVT-297 | Development of a Framework for Validation of Computational Tools for Analysis of Air and Sea Vehicles

Mon, 01 Jan 2018 00:00:00 +0000

In the NATO STO AVT-297 project I was a lead graduate researcher within an international team of experts. My contributions centered on developing a novel systems engineering-based process for validating computational tools used in the analysis of air and sea vehicles. This involved a comprehensive examination of coupled constituent physics and their representation in computational models, assessing their accuracy and relevance for various intended applications. I helped identify a validation database for the intended uses of vehicle systems, with a specific focus on demonstrating the process on a transport aircraft.

The work was a significant part of my Ph.D. research at Georgia Tech and a contribution to systems engineering and model-based systems engineering (MBSE). It advances the understanding of how different physical phenomena interact within simulation environments and how these interactions can be captured and assessed, laying a foundation for more reliable use of computational tools in the design and operation of complex vehicle systems.

The research output of this effort will be published as two chapters of a NATO STO Technical Report.

Smart Campus

Mon, 01 Aug 2016 00:00:00 +0000

Smart Campus is a multi-disciplinary, multi-year, multi-million dollar project that treats the entire campus as a living laboratory, collecting data from energy utility systems. With the abundance of data, it is possible to see improvements in energy performance of building operations with collaborations with GT Facilities Management and draw insights for future building projects. As a senior student in the project, my responsibilities include leading teams of graduate and undergraduate students, helping with writing grant proposals and statements of work, and individual contributions that are listed but not limited to below:

Data Analysis & Anomaly Detection: I have processed and prepared multi-dimensional, structured but poor quality data for various tasks in this project. Problems found in measured data are passed to corresponding professional working for Georgia Tech Facilities Management (GTFM).

Building Energy & Water Modeling: I developed and calibrated detailed time-dependent models in MATLAB Simulink, Dymola and Python of building energy systems based on data obtained from buildings in operation. These models are being used to analyze the impact of various factors and possible extensions on the performance of buildings in Georgia Tech campus. Furthermore, I am lucky to have worked on the amazing (KBISD) since its pre-construction days!

Architecture of the Kendeda Building digital twin (Lewe, 2022).

Weather Data: Weather is one of the primary drivers of building energy performance. In order to make accurate, real-time predictions, the building energy models (digital twins) needs temperature, precipitation, wind speed, relative humidity, and solar irradiance (global horizontal, direct normal, and diffuse horizontal) inputs to predict the thermal performance. This also calls for the use of accurate, local weather data instead of typical meteorological year (TMY) practices. Although Georgia Tech campus has several weather stations, one can come across missing or erroneous data quite frequently. Therefore, other stations in the area are fused to obtain a complete, accurate, and up-to-date weather data set. A neural network was trained using the solar components obtained from the previous four years from the NREL database and applied to estimate diffuse horizontal irradiance where data is not available. The generated set of weather data combines the best-quality data points from the available, local sources so that it can be used for the calculation of the thermal performance of the building. I developed tools and APIs to query weather data from multiple databases (GT & open-source), cleaned and fused them to be used by other entities in the project in real-time.

Continuum Robot

Wed, 01 Jan 2014 00:00:00 +0000

Colonoscopy is a widespread method to inspect the colon for masses or any anomalies and to take a sample of suspicious cancerous tissue with biopsy. This procedure is highly important for early detection of colon cancer and its treatment. During the procedure, a flexible shaft with a camera attached on its tip, is inserted in the colon to monitor or to biopsy. Colonoscopes in the market are guided in the colon with the push of the operator and the tip can be bent by adjusting the tension of cables that lie along the shaft. Since this system has low maneuverability, it can be really hard for the operators to guide it especially through the highly curved parts of the colon such as “sigmoid colon”. The operator has to maneuver through these sections, which can be time consuming or even risky . Possible complications of colonoscopy are; perforation, pain due to mechanical stress and hemorrhage.

At Bogazici University, I worked as a research assistant at the Haptics and Robotics lab under the supervision of Prof. Evren Samur. We came up with a concept for a continuum robot that might be suitable for endoscopy applications. I was the lead researcher in the first two years, mainly at the concept development stage. We formed a multi-disciplinary team with industry experts and medical doctors from Acibadem University to realize this goal. Our proposal to the Scientific Research Council of Turkey was awarded ~$750k.

During my time as a member of this project, I authored a journal paper, presented a poster at ICRA 2015 Soft Robotics workshop, and was awarded a patent valid in Turkey and the EU.

Although I left Bogazici University to pursue a PhD in the US, it is still exciting to see the progress still being made there!