Suggested Projects

These are descriptions of Thesis and Independent Work projects that have been suggested for the 2017-2018 academic year. The topics are organized by major, and quick links are provided here for easy access. To learn more about any topics, please contact the listed professors.




  • Analysis of systems dynamics in bacterial stress responses important for pathogenesis
    To establish an infection a series of host-pathogen interactions need to occur, ranging from quorum sensing and adhesion, to evasion of immune attack and biofilm formation. Antivirulence therapies inhibit the ability of a pathogen to cause an infection by targeting these interactions, and thus represent an attractive addition to our antibiotic arsenal. Nitric oxide (NO) and hydrogen peroxide (H2O2) defense systems are essential for many pathogens to establish and/or sustain an infection, suggesting that inhibitors of these systems would provide potent anti-infectives. Unfortunately, inhibitors of obvious targets in these networks are either toxic to humans or poorly transported into bacteria. In recent years, the Brynildsen lab has generated systems-level models of NO and H2O2 response networks in bacteria, in order to identify additional strategies to sabotage these systems. This project will study the dynamics of these models in order to discover novel emergent properties, and for properties of sufficient interest, experimental confirmation will be conducted. Contact: Prof. Mark Brynildsen.
  • Collective cell dynamics in wound healing
    Contact: Prof. Stanislav Shvartsman.



  • Structural health monitoring and adaptive structures. Contact: Prof. Sigrid Andriaenssens.



  • The role of simplicity in reasoning by machines.
  • Machine learning, including game playing, photo identification, and semantic identification.
  • Applications of machine learning, such as for game playing or semantic classification of spoken utterances.
  • Detecting all objects of some type (such as faces) in a visual scene.
  • Develop a system combining body-mounted cameras and/or Kinect with tactile or auditory feedback to help blind people avoid obstacles.
  • Use computer-controlled milling machines to fabricate bas-reliefs, using substrates of heterogeneous materials.
  • Adapt a MakerBot or other hobbyist-grade manufacturing device to use multiple materials.
  • Modeling of topics from textual sources
    Contact: Prof. Szymon Rusinkiewicz.
  • Improving recognition abilities of computer vision systems (e.g., building models that are better at recognizing objects, human actions, human poses, object attributes, etc. in still images).
  • Building multi-task computer vision models capable of performing several tasks within the same framework.
  • Designing interactive computer vision algorithms that can effectively learn from human feedback.
  • Creating interpretable or explainable computer vision models.
  • Studying the cultural/social/gender bias embedded in existing computer vision systems.
  • Contact: Prof. Olga Russakovsky.



  • Detection and Recognition of Hand Gestures from Video. The goal of this project is to develop algorithms that take as input video from a camera viewing an individual and attempt to recognize some well-defined set of hand gestures made by the individual. This work could potentially have applications to interactive computing and games. Contact: Professor Peter Ramadge.
  • Problems in Machine Learning. Specific applications include feature extraction, clustering, classification, and integration of data collected from experiments in different levels of biological systems or fusion of diversified multimedia data such as text, speech, image, video, and graphics. Theoretical aspects of machine learning include statistical learning, optimization, and algebraic theory. Contact: Professor S. Y. Kung.



  • Baxter the Robot. Baxter is a two-armed assembly robot that is designed to perform dull, industrial tasks. The robot can easily be taught new tasks, and it could be employed in a variety of experiments and applications. For example, using its video camera, it could interface with chess-playing programs, playing against itself or a human player. The robot could be taught to be a surgeon's assistant or, for that matter, to be the surgeon! The robot provides a low-bandwidth platform for positioning high-bandwidth end-effectors, for investigating control algorithms, and for exploring human-machine interactions. It is ideally suited for investigating cooperative two-arm control logic. End-effectors that could be designed and tested using Baxter include compliant grippers, grippers with tactile feedback, pneumatic grippers, micro-grippers for small assembly, octopus tentacles, and solenoid prismatic actuators (as might be used to play pool). Contact: Prof. Michael Littman, Daniel Nosenchuck.
  • Robotic Consumer Appliance. Robotic appliances are beginning to make inroads into the consumer product marketplace. Robotic floor care products, such as vacuums and sweepers have obtained large market shares and revenues. The current project investigates the potential of a robot to significantly aid in the performance of routine daily household tasks. The specific product will be determined by consultation with the advisor. The application will be chosen based on several criteria, including potential consumer acceptance and marketability of the product, and the ability to create a demonstrable prototype within the timeframe of the project. Emphasis will be placed on products that have retail price-points below $100. Particular attention will be focused on human-robot and robot-environment interaction along with intuitive user interface design and autonomous robotic-appliance operation. Products which demonstrate the potential for consumer acceptance will have the opportunity to be subjected to possible commercialization, initially through product test marketing. Contact: Prof. Daniel Nosenchuck.
  • Projects of interest to Prof. Jeremy Kasdin. My group researches technologies for imaging exoplanets and concepts for future NASA missions that incorporate them. We are currently working on the coronagraph instrument for NASA’s upcoming Wide Field Infrared Space Telescope (WFIRST) as well as starshades for that and other missions. Our focus is primarily on control and estimation techniques for managing the optical wavefront in a telescope. I am also always interested in problems in orbital mechanics and spacecraft design and control. I am also beginning a collaboration with Physics on a balloon born telescope that should have good student work opportunities. Some specific possibilities might be:
    • Modeling the surface and mechanical properties of a MEMS deformable mirror in our lab. Work with a graduate student on integrating into our wavefront control approaches.
    • Work with a graduate student on a new wavefront sensor concept for use in WFIRST and implement it in our lab.
    • Work with a postdoc in our occulter test lab to begin implementing a hardware-in-the-loop demonstration of formation flying for an occulter based space mission.
    • Work jointly with a graduate student and Prof. Bill Jones in Physics on a new balloon borne telescope. The current project will involve designing and building a tabetop prototype of the telescope to test alignment and active control techniques.
  • Project of interest to Prof. Luis Gonzalez.
  • Projects of interest to Prof. Michael Littman.
    • Develop a next-generation construction toy set using the MAE Department's new injection molding machine - I have in mind a construction toy set for future architects and builders that one could use to replicate all of the parts that would allow for developing a model framing - the kind of skeletal framing that one sees when houses are being constructed.
    • Improve our Mechanical Analog Computer. The original device was developed by MIT Professor Vannevar Bush and is known as the differential analyzer. The MAE version of the Analog Computer was developed by several students working with Prof. Littman over several years and is made of K'NEX building materials and some 3D printed parts. What is needed is an improvement in performance and speed. The basic computer works for solving first and second order equations, but it is very slow operating (that is, an hour to model a simple transient response). Improvement in speed and accuracy by a factor of ten is the target.
  • Projects of interest to Prof. Naomi Leonard.
    • Projects in my group focus on control of networked multi-robot systems, dynamics of animal groups, bio-inspired design, and human decision-making dynamics. We are interested in the connections between individual dynamics, social interactions and emergent behaviors in animal aggregations such as schools, flocks and herds and control laws for coordinating the behavior of groups of autonomous vehicles for cooperative tasks such as environmental sensing. We are interested in human-in-the-loop systems and we also work at the intersection with dance and music. Undergraduate projects can involve working with theory or working with software and hardware or some of each. We have available an underwater robot testbed and a testbed with wheeled mobile robots, and we are developing new robots.
  • Projects of interest to Prof. Anirudha Majumdar. Our group works on developing approaches for pushing agile robotic systems to the brink of their hardware limits while providing formal guarantees on their safety and performance. I am particularly excited about applying techniques from nonlinear control theory, optimization, motion planning, and machine learning to the areas of autonomous unmanned aerial vehicle flight, legged locomotion, and grasping/manipulation. Some specific possibilities for undergraduate projects could be:
    • Developing a quadrotor with a gripper that can pick up and move small objects.
    • Implementing an obstacle-avoidance controller for a quadruped robot (specifically, the Ghost Robotics "Minitaur").
    • Using techniques from machine learning to create low-dimensional dynamical models of perceptual (e.g., vision) data (with applications to vision-based navigation and control for unmanned aerial vehicles).
    • Implementing a vision-based controller for a quadrotor or fixed-wing airplane in simulation.
  • Project of interest to Prof. Clancy Rowley.
    • Dynamics and control of bicycles. Bicycles provide a rich source of interesting problems in dynamics and control. For instance, a rear-wheel steering bicycle is difficult or impossible to ride, because of the presence of right half-plane zeros in the transfer function from steering angle to tilt angle. Could you design a rear-wheel steering bicycle that you can ride? Alternatively, could you design a stability augmentation system for an ordinary bicycle that helps a beginning rider by improving stability at low speeds? Such a system could be passive, for instance involving the geometry of the fork, or active, involving sensors and motors .
  • Ecology and the Environment 
    • Fundamental Studies of Swimming. We have an ongoing program to study the hydrodynamics of manta ray swimming. We are also interested in building an underwater vehicle (UAV) that swims and maneuvers like a manta ray. We have been challenged by the University of Virginia in a contest to build the best manta UAV. One or more students are sought to take up this challenge. The first competition ended up in a draw, and was reported in Science magazine. We would like to do better. Contact: Prof. Alexander Smits.



  • Automated Ground Vehicles. A vehicle to drive in an urban environment, completely autonomously, avoiding obstacles and finding its own way. There are many ways to contribute to this project, from low-level control of the vehicle, to integrating information from various sensors.
  • Automated highways.
  • Intelligent transportation systems.
  • Contact: Prof. Alain Kornhauser.



  • --.



  • Inference, reasoning, problem solving.
  • Bio-Inspired Control System to Simulate Human Performance. Arguably, humans are equipped with the most advanced motor system on the planet. We possess a remarkable ability to learn new motor skills and retain memories for those skills throughout life, such as riding a bicycle. The ease with which we perform these skills belies their overwhelming computational complexity. At present, we know very little about how the brain performs these computations. Understanding how the brain controls movement will not only help restore function following neurological insult, but also may provide insight into designing better robots and machines. A senior thesis might focus on designing biologically-valid control systems to simulate human movement and skill acquisition. Contact: Prof. Jordan Taylor.
  • Neuroscience
    • Neural Network Modeling of Cognitive Functions. Several members of the psychology department are engaged in the construction of computational models that seek to simulate cognitive function. These are cast at many levels, from symbolic models that attempt to capture higher level cognitive processes such as reasoning and problem solving, to neural network models that target a more detailed understanding of how such processes are implemented in the brain. Contacts: Jonathan Cohen (neural network models of attention, decision making and cognitive control) and Ken Norman (neural network models of encoding and retrieval from long term memory).
    • Brain Imaging of Cognitive Function. Faculty in Psychology, in collaboration with faculty in other departments and the Center for the Study of Brain, Mind and Behavior (CSBMB), are actively involved in the development of new methods for the acquisition and analysis of brain imaging data, and their use for deepening our understanding of the neural bases of cognitive functions. Imaging modalities include functional magnetic resonance imaging (fMRI) and event-related scalp electrical recordings (ERPs), in conjunction with other physiological and behavioral data, such as eye movements, pupilometry, and galvanic skin resistance. These data are used to guide and constrain the development of theories regarding the brain mechanisms underlying cognitive processes such as visual perception, attention, memory, and cognitive control. New methods for studying brain function are under development (such as the integration of virtual reality apparatus with the fMRI scanner), as are advanced methods of image analysis, involving the use of multiresolution temporal techniques (such as wavelet transformations) and blind source separation (such as independent components analysis). Contact: Prof. Jonathan Cohen.
    • Visual perception
      Contact: Prof. Jonathan Cohen.