Squishy Robotics: Machine Learning and Data Science for Deployable Robot-Driven Sensing and Detection of Methane Leaks and Wildfire Onset
Alice Agogino, Professor
Mechanical Engineering
Applications for Fall 2024 are closed for this project.
Description: Squishy robots are rapidly deployable mobile sensing robots for disaster rescue, remote monitoring and space exploration. Our emergent technologies are at the fusion of robotics, mobile sensing, machine learning, big data fusion and smart IoT (Internet of Things).
Our first target market is the HazMat and CBRNE (Chemical, biological, radiological, nuclear and explosive) response market, enabling life-saving maneuvers and securing the safety of first responders by providing situational awareness and sensor data in uncharted terrain. This product can be deployed with multiple – even swarms – of collaborative squishy robots, equipped with visual, audio, chemical, biological, radiological and GPS sensors, that can traverse rough environments and be quickly deployed by ground or aerial vehicle to inform first responders, and assist in the rescue of victims until human first responders can arrive.
Our customizable platform is being designed for dynamically changing situations and we expect future applications to include scientific monitoring, delivery services, smart home appliances, as well as educational applications for K-12 students, teachers, parents, roboticists and hobbyists.
In this project, you will be collaborating with a team of academic researchers - M.Eng and M.DevEng students, postdoctoral researchers, and faculty - along with the Squishy Robotics Team. Your responsibility will involve investigation, adaptation and application of machine learning models to classify and ultimately predict the presence, occurrence, and dynamics of methane leaks and wildfire events. Your work will involve combining multiple scales and sources of data, demonstrating the improvement in model performance based on additional sensors (which would be deployed by Squishy Robots in field applications), and writing scholarly research reports related to your technical work (see below).
Role: Your key work will involve the following:
- Reading, understanding, summarizing, and extracting key actionable insights, models, and codebase/datasets from leading-edge papers in the broad fields of machine learning
- Identifying, gathering, manipulating, cleaning and managing relevant datasets
- Recreating state-of-the-art results in the field of machine learning for climate change (methane leaks) and disaster management (wildfire)
- Collaboratively coding, troubleshooting, applying, and characterizing machine learning models
- Visualizing and reporting results
Qualifications: Students should possess skills in at least one of the following, and strong interest in the others:
Experience with software, coding, and data science skills, especially related to applying machine learning models (e.g., Python, TensorFlow, Google CoLab)
Experience with data wrangling and management
Interest and motivation to read research papers in a variety of fields describing AI / ML models
Data visualization and reporting (e.g. Pandas, R, or other quantitative data visualization interest)
Interest and motivation to address climate change and disaster management via emerging technologies
Day-to-day supervisor for this project: Vivek Rao, Post-Doc
Hours: to be negotiated
Off-Campus Research Site: Weekly mandatory, remote, and synchronous meetings will form the basis of collaboration, and students are expected to collaborate and meet among themselves more frequently. Although much of the work will be to work with a graduate student team in the BEST (Berkeley Emergent Space Tensegrities) Lab on campus, Squishy Robotics, Inc. also has space as a start-up in West Berkeley in new offices.
Related website: http://squishy-robotics.com/
Engineering, Design & Technologies Social Sciences