Master’s Student Projects

Independent/directed Study for Master’s Degree Students

UMTRI researchers have proposed projects that could be used by Master’s degree students looking to gain research experience. We have reached an agreement with the Department of Mechanical Engineering that these projects could be carried through ME590 credits. We are currently in negotiations with other CoE departments as well. Please contact the faculty member for each project to learn more.

UMTRI Project #1: Adaptive Safety Designs for Injury Prevention: Human Modeling and Impact Simulations

Faculty Mentor: Jingwen Hu, 

Project Description: Unintentional injuries, such as those occurred in motor vehicle crashes, falls, and sports are a major public health problem worldwide. Finite element (FE) human models have the potential to better estimate tissue-level injury responses than any other existing biomechanical tools. However, current FE human models were primarily developed and validated for midsize men, and yet significant morphological and biomechanical variations exist in human anatomy. The goals of this study are to develop parametric human FE models accounting for the geometric variations in the population, and to conduct a feasibility study using population-based simulations to evaluate the influence of human morphological variation on human impact responses in motor-vehicle crashes and sport-related head impacts. Specifically, in this study, students will use medical image analysis and statistical methods to quantify the geometric variance of the skeleton among the population; use mesh morphing methods to rapidly morph a baseline human FE model to a large number of human models with a wide range of size and shape for both males and females; and conduct impact simulations with those models toward adaptive safety designs. Prerequisites: Proficiency in Matlab, Interested in injury biomechanics research, Demonstrated ability in FE model development and application is a plus Research Mode: Online or Hybrid

UMTRI Project #2: Data Elements from Video using Impartial Algorithm Tools for Extraction (DEVIATE) 

Faculty Mentor: Carol Flannagan,

Project Description: In everything from autonomous vehicle testing to improving driver safety, video recordings of drivers provide important data, but to be usable, the data must first be extracted. Compared to human coders, automated algorithms have the potential to reduce expense and increase speed, while increasing the accessibility of the wealth of information captured about drivers’ behavior. Our project will focus on developing strategies for automating video data extraction to record vehicle occupant behavior, as well as objects and actions in the vehicle’s forward view. In all of these areas, there is a potential to introduce an unintended bias in the algorithms that could have negative societal implications. Measuring and reducing this bias will be a key goal of our work. Prerequisites: Experience/knowledge of data structures, statistics, algorithm development, human cognition. Experience in Python, R, and/or C++ Research Mode: Online or hybrid

UMTRI Project #3: Driver State Monitoring for Automated Vehicles

Faculty Mentor: Monica L.H. Jones,

Project Description: With increasing automation (SAE Levels 2 and 3), the role of the driver will transition from Driver Driving (DD) to Driver Not Driving (DND). However, there is little quantitative information available on naturalistic non-driving occupant behavior. Driver state monitoring (DSM) systems attempt to predict the driver’s readiness to respond to a takeover request or other emerging need within the situation from information obtained from cameras and other sensors. These systems face several challenges to comprehensively track the continuum of possible driver postures and behaviors.This project will explore the characteristics and behaviors associated with non-nominal postures, driver engagement, monitoring, and state levels (Day vs. Night conditions) that are useful for driver state monitoring (DSM) classification. Continuous measures during in-vehicle test conditions include: subjective assessment, 2D image and 3D depth data, physiological response, and other available DSM outputs. The project seeks to quantify driver response to unscheduled automated-to-manual (non-critical) transitions in an L3 automated driving conditions. The results of this study may identify disallowed states and provide further design guidance for DSMs. Prerequisites: Some experience with scientific programming languages is required (e.g. Mathematica, MatLab, Python), Familiarity with computer vision programming is desired Research Mode: In Lab, Remote, or Hybrid

UMTRI Project #4: Driving Scenario Simulation and Analysis

Faculty Mentor: Shan Bao, 

Project Description: When evaluating and testing automated vehicle technologies, it is pretty challenging and expensive to test the prototype system using real cars on real roads. Ideally, parameter setting of sensors and vehicle control systems can be tested and evaluated under variety of simulated scenarios at first.  This work is sponsored by a mixed of sponsors with several focuses. The work is designed to simulate real world driving scenarios in the virtual environment through software (e.g., Carla) or Virtual Reality techniques. Student interns get to work with the exciting concepts and interact with our industry sponsors directly and will be able to implement your simulation results through hands on experiences. We are looking for multiple motivated student helpers. Training on certain software (e.g., Carla and Carsim) are available. Prerequisites: Have coding experience/knowledge, and are comfortable to work with a big group. Research Mode: Online or Hybrid

UMTRI Project #5: Safety and Independence of Passengers in Wheelchairs Using Automated Vehicles

Faculty Mentor: Kathleen D. Klinich, 

Project Description: We are pursuing multiple projects to ensure that people who travel while seated in their wheelchairs can safely and independently do so in automated vehicles where there may not be a driver to assist in securing the wheelchair. We plan on reviewing relevant standards to identify what manufacturers need to consider when developing accessible AVs, performing testing with volunteers to evaluate usability of different configurations, and develop hardware components to allow independent use of restraint systems.  Prerequisites: Strong technical writing skills, experience with spreadsheet/data analysis, mechanical design/controls experience, and an interest in improving user travel experience and working with people who have disabilities. Research Mode: In lab, hybrid

UMTRI Project #6: Motion Sickness to Inform Automated Vehicle Design

Faculty Mentor: Monica L.H. Jones, 

Project Description: Motion sickness in road vehicles may become an increasingly important problem as automation transforms drivers into passengers. However, lack of a definitive etiology of motion sickness challenges the design of automated vehicles (AVs) to address and mitigate motion sickness susceptibility effectively. The quantification of motion sickness severity and identification of objective parameters is fundamental to informing future countermeasures. Data were gathered on-road and on the Mcity test facility.  Continuous measures include: subjective assessment, 2D image and 3D depth data, thermal imaging, physiological response and vehicle data. Modeling effort will elucidate relationships among the factors contributing to motion sickness for the purpose of generating hypotheses and informing future countermeasures for AVs. Prerequisites: Some experience with scientific programming languages is required (e.g. Mathematica, MatLab, Python), Familiarity with computer vision programming is desired Research Mode: In Lab, Remote, Hybrid

UMTRI Project #7: Development and Assessment for the Automated Overtaking Feature

Faculty Mentor: Brian T. W. Lin, 

Project Description: Cyclists share the roadway with motor vehicles that drive much faster. Once an accident occurs with bicyclists involved, the death rate of the bicyclist is extremely high. This study follows a systematic method to develop a prototype for an automated overtaking system, specifically for overtaking bicyclists. Naturalistic driving data based on pre‐extracted overtaking events with more other critical factors will be mined to create three models that covers the four phases of an overtaking: approaching, overtaking, passing, and returning. These models will then be implemented as an automated overtaking prototype to a simulated platform for a motor vehicle to overtake bicyclists based on different strategies. An experiment of human study will be conducted to evaluate the prototype from both the viewpoints of the driver and the bicyclist that how they want to overtake and be overtaken safely. It is expected that the outcomes can offer the OEMs and suppliers who are keen on developing safe and human‐centered automated vehicle systems with useful insights. Furthermore, the insights can be helpful for legislation on the act or guidelines of protecting on‐road vulnerable bicyclists. Prerequisites: 1) will help conduct evaluation methods, collect subjective assessment data in a driving simulator, conduct data analytic methods, and analyze the data with statistical methods 2) must have taken the classes of statistics, and human factors or usability assessment 3) familiar with MATLAB or R to analyze large datasets Research Mode: Remote, Hybrid

UMTRI Project #8: Body Dimension Estimation from Clothed 3D scan

Faculty Mentor: BK-Daniel Park, 

Project Description: Three-dimensional (3D) surface measurement has become a central component of anthropometric surveys. Modern surface scanning equipment can accurately capture the shape of the surface of the body in a few seconds. However, the practical aspects of conducting 3D scanning surveys have changed little in the past decade. In particular, participants are required to change into close-fitting garb that minimizes the clothing effects on the subsequent scan. This clothing ensemble must be provided, along with suitable privacy for changing, and the consequence is that several seconds of scanning can require 10 minutes or more of preparation and considerable resources. The University of Michigan Transportation Research Institute (UMTRI) recently introduced a new body shape estimation method, “Inscribed Fitting” (IF), that is much faster than previous techniques, requiring at most a few seconds of computation on a typical computer. This IF method uses an iterative process to estimate the body shape underlying the clothing, based on the observation that the correct body shape is the largest body shape that does not protrude through the clothing. The main objective of this study is to develop a standalone software system that can be used to estimate body shape, standard anthropometric dimensions, and body landmark locations from scan data obtained from individuals in arbitrary clothing ensembles. Prerequisites: Proficiency in computer programming languages (C#, Python, etc.), Interested in computer vision/machine learning research, Excellent oral and written communication skills, Ability to work well as a member of a team Research Mode: In Lab, Remote, Hybrid

UMTRI Project #9: Identify and Testing User Interface Design Needs for AV-VRU communications

Faculty Mentor: Shan Bao,

Project Description: For the vulnerable road user community and individuals (VRU, i.e., pedestrians and bicyclists), effective communication between Automated Vehicles (AV) and VRUs is crucial to VRUs developing trust as they interact with AVs. To support the effort of improving mobility and safety for all Americans,this study is d signed to address this issue. This project will collect user needs data and assess the information display efficiency and system suitability of accommodations to address common needs by using the principles of user-centered design to ensure that the prototypes have been developed to maximize the use of AVs by VRUs. By working on this project, you will have the opportunity to work on an exciting research topic and get connected with experts from leading companies in this domain. Prerequisites: Team players who are motivated in working with other group members. Experience with conducting study to collect data from human participants are plus! Research Mode: Online or Hybrid

UMTRI Project #10: A Tool for Augmented Reality (AR) Assisted Surgery: 3D Human Modeling and Visualization

Faculty Mentor: Jingwen Hu,

Project Description: An AR-assisted surgery tool will provide a composite view between computer-generated patient anatomy and a surgeon’s view of the operative field, which may lead to more precise understanding of the detailed anatomy and also significantly increase accuracy in tumor localization and resection. In this study, we will focus on a software tool that can address the rapid development of computer anatomy models and accurate registration between the anatomy model and real patient geometry, which are the two key aspects of AR-assisted surgery tools.  We plan to use an AR device, Microsoft HoloLens, as the main hardware to demonstrate the software capability, although our software should not be limited to HoloLens only.  In this study, we will use liver surgery as an example, thus the medical images and anatomy models will only focus on the liver and the surrounding tissues.  Because liver is the largest solid organ in the abdomen, is pliable, and operative interventions can alter its anatomy, it will pose significant challenges on model registration, which will be a good test for the AR-assisted surgery tool. For surgeons who have to deal with complex anatomical structures that are not always visible, the proposed AR-assisted surgery tool will provide much needed understanding of anatomic relations beneath the surface, and will likely lead to better accuracy, safer resection, lower complications, and superior surgical outcomes. Prerequisites: Proficiency in computer programming languages (C#, C++, Python, etc.) Research Mode: Online, Hybrid

UMTRI Project #11: Development of a realistic simulation environment to train and test autonomous vehicles

Faculty Mentor: Arpan Kusari,

Project Description: An essential component of the training and testing of autonomous vehicles rests on having a simulation environment capable of producing vehicular data which rivals the complexity of real-world scenarios. However, the majority of current simulation environments utilize traffic flow simulation in order to generate and operate the surrounding vehicles which results in a huge gap between the simulation environment and actual driving. We propose to develop an open source  simulation environment by incorporating human driver  behavior models and subsequent macroscopic intentions into the simulation environment. This simulation environment would utilize the naturalistic driving data already present at UMTRI in an “intelligent” manner to seed behaviors in the surrounding traffic. Research tasks for students include: Research into the most popular traffic simulation environments as background (10%), Create a simple simulation loop with basic motion models and ACC behavior for other vehicles with Python as front end and C++ as backend (40%), Utilize extracted naturalistic driving data to seed traffic vehicle behavior (20%), Add complex behavior for other vehicles utilizing intentions (20%), Build/utilize other complex map layouts (10%) Prerequisites: Expertise in Python, and C++ is required; knowledge of robotics is desirable Research Mode: Online or hybrid

UMTRI Project #12: State of health monitoring and transmission accuracy assessment for road-side equipment employing Connected Vehicle Technology

Faculty Mentor: Jim Sayer,

Project Description: Over 75 Connected Vehicle devices are deployed at intersections, curves and mid-blocks crosswalks around Ann Arbor.  These devices can transmit information to approaching vehicles about the color and timing of traffic lights, or the presence of a pedestrian in the road ahead.  Of interest to researchers is the operational state of the Road-side Units (RSUs), as well as the accuracy of the transmissions being broadcast.  Currently, the signals are only used to provide warnings to drivers, but in the future automated vehicles may rely on accurate information being generated by these road-side units to navigate intersections safely. In this study, students will use custom designed software and hardware while in the field to test the state of RSU’s around Ann Arbor and report the results.  Students will also be involved in helping to properly configure devices that are not properly functioning.  Work may also include tailoring pedestrian detection devices to more accurately detect crossing pedestrians, and tailoring in-vehicle warning parameters to best provide information to drivers about potential hazards including red-light violations. Prerequisites: Driver’s License, Interest in Connected Vehicle Technology Research Mode: Online data analysis with local driving around Ann Arbor (vehicle provided)

UMTRI Project #13: Pedestrian Detection Using via Ultra-Wideband (UWB) Device for Smart Vehicles

Faculty Mentor: Henry Liu,

Project Description: Pedestrian detection is one of the critical problems in the autonomous driving field. Traditional methods are either unreliable, e.g., a camera-based detection system, or expensive, e.g., a lidar-based detection system. As in 2019, Apple starts to put Ultra-Wideband (UWB) functionality into the iPhone; it is expected that shortly, most smartphones and other smart devices carried by humans will be equipped with UWB functionality. As the UWB device will give accurate localization results, it becomes an economical and precise method for pedestrian tracking and collision warning for smart vehicles. This project will develop a data-driven approach that detects and localizes pedestrians via UWB localization. If interested, the student can develop a cellphone app to illustrate the developed method. Prerequisites: Proficient programming skills, basic mathematics. Research Mode: Hybrid (online)

UMTRI Project #14: Modeling Spatial Autocorrelation in Travel Patterns Influenced by Household Level Exogenous Events 

Faculty Mentor: Aditi Misra,

Project Description: Travel related decisions are influenced by multiple factors at personal and household level and have been studied extensively using decision theories. For example, preference of travel mode is a personal choice, but often restricted by household level availability of vehicles, leading to development of joint decision making models of trips and mode choices for individuals within a household. However, trip and mode choices are also influenced by factors beyond the household and at community level that have not received similar attention. In this project, we will test the hypothesis that (i) people’s travel choices are influenced by the community they are living in and (ii) people self select themselves into living in certain neighborhoods with certain characteristics that align with their values. Natural disasters and disruptions are examples of events where these spatial influences are most pronounced – people evacuate when they see neighbors evacuating. The experience with and data from travel or lack thereof during COVID 19 provides a natural experimental setting to test this hypothesis and in developing generalized models addressing these biases that can later be extended to other extreme event scenarios. Prerequisites: Required: Strong background in Statistics, particularly with different families of Regression/ Causal Inference Models with some understanding of Time series modeling; Ability and experience with modeling in R and/or Python Desired: Geospatial data analysis skills; Interest in travel behavior and decision theory Research Mode: Online, Remote

UMTRI Project #15: Development of Standard to Define Driving Performance Measures and Statistics

Faculty Mentor: Paul Green,

Project Description: The faculty mentor has written an automotive engineering practice (SAE J2944 – Operational Definitions of Driving Performance Measures and Statistics, 171 pages) that defines measures such as standard deviation of lane position, lane departures, steering entropy, etc., that are cited in publications on how people drive.  That document is primarily limited to most wheeled vehicles driven on road.  The goal of this project is to update that document based on research that has occurred in the last 5 years, and to extend it to include off-road driving and tracked vehicles.  The end result will be a military standard that in theory should cited in every study of driving.  What is needed are mini-literature reviews to move the document forward.  This is a project of some size, and at least 6 other people will be working on it. Prerequisites: skills in reading technical literature, Licensed driver Research Mode: In person, Hybrid

UMTRI Project #16: Using the Unreal Game Engine for Collecting Driving Performance and Workload Data

Faculty Mentor: Paul Green,

Project Description:  The Driver Interface Team has been conducting research with others for the Army on the remote driving of vehicles.  This has been done using the Unreal game engine, currently in the Yehorivka virtual world, but others are of interest.  At this point we are making a number of modifications to the Unreal engine and/or software connected to it, to make it suitable for research.  They include being able to:  (1) record data from a variety of input devices (steering wheel/ pedal assemblies, game pads, yokes, etc.), (2) supporting a wider array of display devices (head-mounted, multiple screens), (3) supporting viewports (forward scene and map), (4) being able to switch between (and record) input devices in real time, (5) being able to periodically record the forward scene for studies of visual demand, (6) providing a wider variety of vehicle dynamics for off-road vehicles, (7) being able to load in other virtual worlds, and (8) collecting baseline driver for young drivers operating these simulations.  We have someone working on this now and are looking to adding to the team. Prerequisites: Knowledge of C++, Python, and Java are desired but not required Experience in creating video games or driving simulations is desired Research Mode: In Lab, Hybrid

UMTRI Project #17: Enhancing CARLA-based Driving Simulations for Human Factors Research. 

Faculty Mentor: Paul Green,

Project Description:  The Driver Interface Team is leading a Multidisciplinary Design Program (MDP) project to enchance the CARLA driving simulator for use in a class on automotive human factors and to facilitate its uses for studies of driver workload, driver distraction, and the operation of partially-automated vehicles.  To date, at team of about 12 people has been working on this project for just over a year.  There 3 current activities within that project.  First, we are developing a graphical user interface that will allow nonprogrammers to create experiments involving traffic in (1) simple urban situations and (2) on expressways.  Second, we are creating a virtual, drivable North Campus that can be loaded and driven.  Third, we are working a connecting a simple motion base (for a vehicle cab) and having it respond to pitch and roll motions produced by the simulations.  As we make progress on these improvements, we need to collect baseline data for people driving the enhanced simulation. Prerequisites:  Knowledge of C++, Python, and Java are desired but not required Experience in creating video games or driving simulations is desired, or CARLA Research Mode: In-person, Hybrid

UMTRI Project #18: On-Demand Delivery Estimation and Routing 

Faculty Mentor: Tayo Fabusuyi,

Project Description: Driven primarily by the Covid-19 pandemic, the volume of online transactions has increased dramatically. This is noticeable with the demand for online product deliveries such as groceries, store packages and restaurant deliveries. While this presents business opportunities, we lack publicly available delivery data, making it difficult to prototype delivery services for cities. Currently, there are no existing methodologies for solving this problem – thus the impetus for the current research project. Our approach will be to enrich the publicly available data with proprietary metadata that has information on the nature of the product ordered, the cost and the time the order was made and anonymized delivery addresses. Routing of vehicles for making the deliveries will be implemented by using a variant of the vehicle routing problem (VRP), modified to reflect the time window and the equity considerations of the demand for online package deliveries. Prerequisites: AI and Data Science skills (specifically raking and routing algorithm design and implementation) Research Mode: Online, Remote.

UMTRI Project #19: Deep graphical approaches to 3D LiDAR data modelling

Faculty Mentor: Arpan Kusari,

Project Description: There has been a proliferation of laser scanners and 3D cameras in domains as varied as mobile robotics , autonomous vehicles and advanced manufacturing along with traditional domains such as land surveying. These 3D scanners output discrete point samples of the underlying surfaces. Thus, the deep learning approaches initially developed for regular camera imagery suffer from sub-optimal performance when transferred to LiDAR point cloud data. We can model the LiDAR point cloud data as a neighborhood graph which can serve as a function of the underlying surfaces. Deep graphical models (DGM) have provided great results  on graphical structures such as citation datasets and genomics datasets. In this project, we would like to explore DGM to LiDAR data modeling specifically1) Research into the most popular DGMs as background 2) Large scale noise removal for LiDAR data using DGM 3) Classification of primitive shapes using DGM Prerequisites: Experience in AI/ML algorithm development Research Mode: Online or hybrid

UMTRI Project #20: Re-constructing pedestrian crash-injury cases through MADYMO modeling and machine learning

Faculty Mentor: Jingwen Hu, 

Project Description: With the recent proportion increase of pedestrian injuries among crash-induced injuries, the current pedestrian safety studies in the U.S. continue to rely on pedestrian injury data collected two decades ago.  The increase of vehicles with automated driving systems will also affect the pedestrian impacts.  Therefore, it is critical to generate a new pedestrian injury dataset to reflect pedestrian injuries in recent years and better evaluate the effectiveness of future safety technologies related to pedestrian protection. The objective of this proposed study is to use computer-simulated crash-injury data and machine-learning algorithms to automatically reconstruct a large number of pedestrian crashes based on linked police-reports and trauma data. Prerequisites: Basic understanding of rigid-body-based modeling (Madymo will be a plus), Experiences in AI/ML algorithm development, Basic understanding of injury assessment will be a plus Research Mode: Online or hybrid