Human-Robot Teaming
Mentor: Dr. Julie Adams

This project seeks to develop new methods for adapting interaction between a human and a robot in a one-to-one partnership relationship based on the human's cognitive workload. The human dons various wearable sensors that provide a data stream that is analyzed to detect high and low workload conditions. Once a high or low workload condition is detected, the robot can adapt its interaction method with the human. REU students will work to develop new algorithms for detecting changes in workload and new algorithms for adapting the robot's interactive capabilities based on the workload detection. The project will require algorithm design, implementation and software testing. Further, REU students will evaluate the algorithms with actual human-robot teams.

Mentors: Dr. Cindy Grimm and Dr. Ravi Balasubramanian

Humans have no trouble picking up and manipulating objects, yet we've struggled to impart that ability to robotic manipulators. This is in part because robotic manipulators lack much of the sensory feedback human hands have, but it's also because we, as humans, are not very good about reasoning about what we do instinctively. The goal of this project is to develop tools and user studies that will let us capture that knowledge and apply it to robotic manipulation tasks.

Example projects:

  • Design (and evaluate) novel interfaces for specifying robotic grasps suitable for non-technical users
  • Determine how perceptual and shape-based cues influence how humans specify grasps
  • Apply statistical and machine-learning based approaches to gathered data in order to specify novel grasps


Design of Implants for Attaching Muscle and Tendons to Improve Human Hand Function
Mentor: Dr. Ravi Balasubramanian

Current reconstructive orthopedic surgeries use sutures to attach muscles and tendons. However, this leads to poor surgical outcomes because of the suture’s limited ability to transmit the muscle’s forces and movement to the tendons. It is expected that using passive implants, such as pulleys and rods, to surgically construct mechanisms in situ using the existing biological tendons will significantly improve post-surgery function (when compared to using sutures) and lead to the development of new surgical procedures. 
Example student projects:

Privacy and telepresence
Mentors: Dr. Cindy Grimm and Dr. Bill Smart

As robots become more ubiquitous there are open questions about how we should think of robots in our living spaces. In this project we look specifically at telepresence - "skype on a stick" - where a remote user controls a semi-autonomous robot in your home or office. Unlike a person in your space, we can control what the robot can see, do and hear - for example, blurring or blacking out part of the video to hide what's there, or preventing the robot from going into some rooms.

This project takes a broad look at privacy in the context of remote presence systems, drawing from related work in areas from legal studies to psychology. We are trying to figure out how people think about privacy in the context of these systems, how we can build interfaces that allow people to specify their privacy concerns, and how we can build the underlying technology that supports privacy-protection.

Example Projects:

  • Interfaces for specifying privacy concerns for remote presence systems.
  • Real-time modification of sensor streams (camera images) to protect privacy (by redacting, blurring, replacing, etc)
  • Setting and enforcing physical restrictions on remote presence systems

Mentor: Dr. Ross Hatton

Spiders use silk as a tool for moving through space and capturing prey. We're building robots that use these principles to move through complex spaces and lasso targets.
Example Project:

  • Controlling the motion of a small legged robot hanging from a rope.

Mentor:Dr. Ross Hatton

Spiders use vibrations in their webs to sense the presence of other insects, similarly to how whales, bats, and submarines use sonar to hunt their prey. In collaboration with a biology team, we're building and instrumenting a giant spider web to examine how this works.
Example Project

  • Designing new test scenarios (how does web construction affect signal transmission) and building the equipment to collect data from the web. We are also designing equipment to shake and pluck real spider webs and compare their responses to our one.

Learning from Humans for Robotic Deburring

MentorsDr. Ravi Balasubramanian and Dr. Burak Sencer

Burrs are undesirable projections of material at the edges of a finished part’s surface. They pose a fundamental problem for manufacturing operations since they affect part handling and assembly operations and lead to part failure. Deburring, the process of removing burrs, is currently performed by human operators. But deburring is labor-intensive and causes injury to operators due to the forces and movements involved. This project seeks to learn the deburring process from a human (including forces and movements involved) and develop a robot to automate the deburring process.
Example student projects:

  • Collect and analyze data (motion-capture kinematic data, force data) of a human operator performing the deburring task
  • Develop algorithms for a robotic manipulator that produces similar movements and forces for deburring

Autonomous Research Vessels to Explore Extreme Ocean Environments

Mentor: Dr. Geoff Hollinger

There are many ocean environments which are unsuitable for manned research vessels, either because they are too dangerous (e.g., near a calving glacier or in the deep ocean), or require too many resources to be effective. We seek to design a new generation of Autonomous Research Vessels (ARVs) that can be programmed to measure ocean dynamics in extreme environments. This research project involves the hardware design, construction, and programing of robust surface and underwater systems that will be used to explore ocean dynamics in Greenland, Alaska, and in remote ocean basins. The REU student will join an interdisciplinary team of researchers from mechanical engineering and oceanography to assist in building and programming an ARV currently under construction at Oregon State University.
Example projects:

  • Programming autonomous surface craft to operate without operator control
  • Simulator design for underwater vehicles to move in the deep ocean
  • Learning from human operators to coordinate autonomous marine vehicles


Autonomous Mapping with Teams of Aerial Rotorcraft
Mentor: Dr. Geoff Hollinger

Autonomous aerial vehicles (e.g., quadcopters) have increased in prominence over the last few years and are seen performing aggressive maneuvers and manipulation tasks on YouTube. These vehicles typically cost upwards of several thousand dollars with a full sensor suite. An open question is the design of a lower-cost vehicle capable of performing autonomous exploration and mapping tasks. This research project involves the development of a team of low-cost rotorcraft capable of fully autonomous exploration and mapping without pre-installed infrastructure. The REU student will primarily assist with implementing the software architecture for the rotorcrafts and testing the flight control, coordination, and mapping accuracy during autonomous operation.
Example projects:

  • Testing the accuracy of the indoor localization using a Kinect sensor
  • Simulation studies of multi-robot coordination algorithms for exploration
  • Programming navigation capabilities for aerial vehicles

Passive Dynamics and Applied Control for Legged Locomotion

Mentor: Dr. Jonathan Hurst

Students on this project will participate in research with our human-scale walking and running robot Cassie. The research requires an interdisciplinary approach combining mechanical design, software, and electronics. Interns may assist with any aspect of this work, focusing on their individual areas of talent, such as mechanical design in SolidWorks, building robot parts, electronics design or wiring, microcontroller software development, Linux-based or Matlab- and Simulink-based software development.


Example Student Projects:

  • Create several prototype robot feet, and advise on control methods to incorporate foot.
  • Help with testing and experimentation of Cassie outdoors and indoors.
  • Begin initial design effort for new arms for Cassie 3, in cooperation with graduate students. 
  • Incorporate decision-making algorithms with real-time robot dynamics, to enable footstep planning and balancing at the same time.

Biologically-inspired Soft Robots
Mentor: Dr. Yigit Menguc

Biological tissues have rich structural and material compositions that give them elegant mechanical properties, such as the smooth variation in elasticity of a tendon that allows it to connect muscle to bone. We will explore patterning hard and soft materials with a multimaterial 3D printer to create structures that can bring biologically-inspired mechanical improvements to soft robots. For instance, one of the challenges in building a soft, squishy robot is constructing durable interfaces between extremely compliant materials such as silicone and rigid parts like batteries. 
Example Projects:
  • Designing heterogeneous structural or material patterns and characterizing their mechanical properties
  • Investigating and implementing algorithms to generate structures inspired by biological tissues
  • Developing a tool to model or simulate soft robotic components
  • Designing, fabricating, and testing a bioinspired multimaterial soft robot

Control of a self-driving wheelchair
Mentors: Dr. Bill Smart

We are working on a self-driving powered wheelchair, for use by people with severe motor disabilities, such as Amyotrophic Lateral Sclerosis (ALS, or Lou Gehrig's disease) and quadriplegia. The goal is to develop a low-cost package that can be added to a traditional powered wheelchair, and turn it into a self-driving system. Self-driving capabilities would afford the wheelchair user more independence, and allow them to move about their world more easily and efficiently.

The project leverages the Robot Operating System (ROS) software for navigation and localization of the system, with special-purpose code for wheelchair-specific features. Our goal is to release both the hardware design and software for the system under an open-source license.Our goal is two-fold: understand how people think about privacy and how that is the same (or different) when it is a robot and not a person, and develop algorithms that lets us "try out" various methods for preserving privacy.
Example Projects:

  • Design, implementation and testing of wheelchair specific navigation and localization algorithms
  • User interface design, implementation, and testing for the wheelchair system.
  • Long-term autonomy issues: map building and maintenance, data logging and management, and learning from experience.
  • Design, implementation, and testing of wheelchair-specific autonomous behaviors.

 Multi-Robot Coordination

Mentor: Dr. Kagan Tumer

Many interesting real world problems require multiple robots to work together. For example, search and rescue missions require coordinating dozens of autonomous robots, as well as ensuring that the robots and humans work together. But providing hard-coded coordination instructions is too limiting. This project explores the science of coordination, and focuses on how to provide incentives to individual robots so that they work collectively.
Example Projects

  • Programming intelligent decision making for robots
  • Implementing incentives for robots
  • Testing coordination algorithms in hardware (wheeled and flying robots)

Bat Sonar for Robots

Mentors: Dr. Cindy Grimm and Dr. Yigit Menguc

Bat sonar is an incredibly rich sense - bats can spot and catch insects in the air, fly through dense forest canopy - all by emitting chirps and listening for echos. They accomplish this feat through deformable ear and nose geometries that let them "shape" the sound scape. In this project we aim to build a soft, deformable bat ear and nose in order to mount such a sensor on the robot.

Example projects:

  • Design a soft actuator to bend the ear and nose
  • Experiment with different geometries to see how they shape sound
  • Apply machine learning to map the returned sonar signal to "object detected" in order to let a robot navigate through a space