MIT Beaver Works Summer Institute

Massachusetts Institute of Technology

MIT Lincoln Laboratory

MIT Lincoln Laboratory Beaver Works

MIT Beaver Works Summer Institute The MIT Beaver Works Summer Institute (BWSI) is a rigorous, world-class STEM program for talented rising high-school seniors. In its first year, the program featured an intensive, hands-on, project-based challenge through which the students learned how to navigate an autonomous mini-racecar through a complex environment. This program culminated in a final race day when the students put their new skills and knowledge to the test against the clock. The first year of BWSI was a complete success thanks to the enthusiasm of our students, the dedication of our instructors, and the hard work from our team members. The Institute partnered with 24 high schools across the nation to recruit the future engineers who participated in our program. We had the pleasure of working with 46 students, 24 of them joining us from outside Massachusetts. Next year, we hope that early-adopting high schools will implement our program’s courses for their students and join the final race day competition. These courses are designed to be just the beginning of the national and then international expansion of engineering education we want BWSI to spark. In the coming years, we will b e adding new BWSI programs in a range of disciplines to increase participation substantially and to build a model that high schools and universities can emulate. While creating a network of these institutes across the nation and the globe, we will support high school STEM teachers who want to use our teaching materials to help better prepare their students for college and potential future careers. We want this network of institutes to work together to bolster engineering education nationally and internationally and provide more opportunities for students interested in STEM fields. Thank you for your continued support of our program.

What is Beaver Works? Beaver Works is a joint venture between MIT Lincoln Laboratory and the MIT School of Engineering that is envisioned as an incubator for research and innovation. Beaver Works facilitates project-based learning, a hallmark of an MIT education, and leverages the expertise and enthusiasm of MIT faculty, students, researchers, and Lincoln Laboratory staff to broaden partnerships with MIT campus and students. The Beaver Works center located in Cambridge, Massachusetts, provides these facilities: areas for collaborative brainstorming; workshops and tools for fabricating prototype systems; and space for classroom-style instruction. Beaver Works allows students to address real-world problems and issues, engages students in hands-on learning, and demonstrates an effective strategy for teaching complex engineering concepts. Beaver Works supports MIT student involvement in a range of research and educational pursuits, including two-semester, course-based capstone projects; joint and individual research initiatives; and Undergraduate Research Opportunities Program internships. Students involved in these projects develop innovative solutions to real-world problems and gain an exceptional experience in hands-on learning from world-class researchers. In addition to the Summer Institute, Beaver Works is also extending project-based learning opportunities to local K–12 schoolchildren. Among these offerings have been a robotics workshop for an all-girl FIRST (For Inspiration and Recognition of Science and Technology) LEGO League team, a hands-on camera-building activity for high-school girls, and a one-day workshop on radars for students in middle school.

2016 Beaver Works Summer Institute Program Summary

Mini Grand Prix Challenge The MIT Beaver Works Summer Institute is a four-week residential STEM program for talented rising high-school seniors. The program in 2016 was the Autonomous Mini Grand Prix, a hands-on, intensive, project-based challenge that demonstrated autonomous mini racecar navigation in a complex environment.

Nine teams of five students each used their own RACECAR (Rapid Autonomous Complex- Environment Competing Ackermann-steering Robot), complete with sensors and an NVIDIA computer running on ROS (Robot Operating System), to… Move!…Explore!…Learn!…Race!

Sparkfun

2D Lidar Scanner

1/10 Scale Electronic Speed Control

In weekly modules, students learned, designed, and implemented algorithms and software from several exciting areas of engineering—motion control, vision-based detection systems, mapping, and autonomous navigation—to demonstrate the advanced capabilities of a self- driving car, with interactive technical coaching from associate instructors. They also received instruction in effective collaboration and communication in a fast-paced, interdisciplinary engineering environment. Every week in mini-challenges, the teams demonstrated the software that they designed and implemented, leading to the final challenge, when the team that completed the course the fastest was crowned the Autonomous Mini Grand Prix Champion! The main program units were: Week 1: Move Basic motion control and simple obstacle avoidance Week 2: Explore Vision-based blob, target, and object detection Week 3: Learn Mapping, localization, and road network navigation Week 4: Race All pieces come together for the final Mini Grand Prix Challenge!!

Class of 2016

2017 MIT Beaver Works Summer Institute (July 10 – August 4, 2017)

Mini Grand Prix Challenge

Autonomous Air Vehicle Racing

Cog*Works: Build Your Own Cognitive Assistant

Adjustable course gates

Reconfigurable obstacles

Safety net

Course floor outline

Mini Grand Prix Challenge (2017) The Summer Robotics program is being expanded in 2017 because of the success of the previous residential program. This program consists of three components: (1) online course from December 2016 to May 2017, open to all interested and committed students;(2) four-week summer program at MIT for a small group of students, July 10–August 4, 2017; and (3) a final competition open to all qualified teams, August 4 and 5, 2017. The 2016 curriculum is being made available online as a series of self-paced units for students to follow independently. This content will become the prerequisite material for participation in the 2017 Summer Robotics program, which will be taught at a higher level and will address more research-oriented technologies, techniques, tools, and applications. For the 2017 summer program, we anticipate more in-depth coverage of pose estimation and recovery for localization, visual servoing for local navigation, machine learning with neural networks for object detection and identification, simultaneous localization and mapping (SLAM), and global planning for navigation in unknown and dynamic environments. For the online course, we plan to use MIT’s EdX platform to prepare the materials that will be linked to content, lab exercises, and demonstration challenges. The online course will provide familiarity with the program’s key tools, such as the Robot Operating System (ROS, http://ros.org) running on the Ubuntu distribution of the GNU/Linux operating system. The online course will develop programming skills with the Python programming language and will provide students valuable experience with the OpenCV (open computer vision) tools widely used in industry and academic research. In addition, the available core units in control systems, computer vision, localization, navigation, and planning will closely follow the learning sequence of the 2016 curriculum. The designs and list of components required to build your own RACECAR are available on GitHub (at http://github.com/mit-racecar) for those who would like to begin now. To get started immediately, a virtual model of the RACECAR that can be controlled through ROS and is suitable for use in the Gazebo simulator is also available in the downloadable materials for the course. To facilitate the build, a virtual machine image with the basic components required for the course is also available for download. This virtual machine is suitable for use with the free, open-source VirtualBox (http://virtualbox.org) or the commercial VMWare Player/Fusion (http://vmware.com) applications. The Mini-Grand Prix Challenge is also being expanded in 2017 to include high school teams from around the country that have adopted the RACECAR platform and utilized it for a robotics program in their own departments or clubs. A new, more challenging venue is being prepared for the first Invitational Mini-Grand Prix Challenge to be hosted on the MIT campus in summer 2017.

RACECAR Online Course Outline

The online course utilizes the OpenEdX platform. Content will be released in stages as it is developed and should be considered under continuous construction. Check back often as we anticipate frequent updates to content based on feedback received from students and incorporation of significant new material as it is released.

Module 1: Basic Prerequisites (January) • Virtual machine installation and use • Basic Linux and command line tools • Introduction to Python programming • Robot Operating System (ROS) tutorials

Module 2: Simulation in ROS (February) • Introduction to the GAZEBO simulator • RACECAR model capabilities and interfaces • Controlling the RACECAR • Visualizing RACECAR sensor data

Module 3: Basic Control Systems (March) • Open loop control • Simple feedback control • Proportional-Integral-Differential (PID) control • Wall following controller

Module 4: Fundamentals of the OpenCV library (April) • Image formation and formats • Color representations • Image segmentation and blob detection • Visual servoing, RACECAR line/lane following

Module 5: Navigation and Planning Concepts (May) • Localization and pose estimation • Global planning and mapping • Local planning in dynamic environments • Potential Field Controller

Autonomous Air Vehicle Racing (2017) In a new offering in 2017, students may participate in an experimental program to build and race autonomous air vehicles. First-person-view drone racing is a new and rapidly expanding sport, but completely autonomous air vehicle (drone) racing is also becoming possible. In this pilot program, students will build, program, and race an autonomous flying drone through an aerial race course. These students, working with advanced sensors and processors and implementing computer vision techniques, will demonstrate and perhaps even innovate the cutting edge of airborne autonomy. This program consists of two components: (1) an online course from January to May 2017, open to all interested and committed students and (2) a four-week summer program at MIT for a small group of students from July 10 to August 4, 2017. All students will learn how to use the software development environments that are a part of the summer course through a series of prerequisite online exercises from January to May 2017. The skills they develop will be demonstrated on a small toy quadcopter so that they can rapidly test their concepts on a course that is forgiving of crashes. Students will be able to demonstrate basic vision-based autonomy after completing the exercises. Students who have successfully completed the online course will be considered for participation in the summer program. Assisted by associate instructors, teams of students will learn how to construct, program, and fly autonomous racing drones. Every week ends with a mini-challenge in which teams will demonstrate their advanced platforms and software build, and will compete for points (and bragging rights!). The program culminates in a final challenge—an autonomous drone race through a 3D race course. The team with the most points will be crowned the MIT Autonomous Air Vehicle Champion!

Conceptual Autonomous Air Vehicle Race Course • Adjustable, highly visible obstacles • Easily reconfigured • Forgiving to impacts (few hard surfaces) • Drives vision & navigation autonomy • Netted area for viewer safety Adjustable course

gates

Reconfigurable obstacles

Safety net

Course floor outline

Autonomous Air Vehicle Racing Course Outline The Beaver Works Summer Institute Autonomous Unmanned Aerial Vehicle (UAV) course will have an online prerequisite component designed to prepare students for the four-week course. Students must complete this course before they arrive on campus this summer. This online curriculum will contain the foundational knowledge required for learning the techniques and algorithms that will be covered during the BWSI course. Online students will learn the basic mathematical principles and computer skills required for designing algorithms and writing software for autonomous quadrotor platforms. After these basic lessons are completed, the online material will cover specific skills required for the autonomous navigation of a quadrotor. This material will be reinforced through the use of an online quadrotor simulator and other online exercises. At the completion of this online curriculum, students will be prepared to complete the hands-on exercises and projects that will be presented during the four-week BWSI Autonomous UAV course.

Module 1: Introduction to Quadrotors • Why Quadrotors • Principles of Flight • Quadrotor Research

Module 2: Introduction to Linear Algebra • Introduction to Vectors – Sections 1.2 – 1.4 • Linear Transformations – Sections 2.1, 2.2 and 2.4

Module 3: Introduction to Matrix Manipulation • Operations on Matrices – All • Matrix Multiplication – All

Module 4: Introduction to Probability and Statistics • Introduction to Probability – Lecture 1 • Conditioning and Independence – Lectures 2 and 3 • Discrete Random Variables – Lecture 5 • Continuous Random Variables – Lecture 8

Module 5: Computer Programming Fundamentals • Linux Introduction • Linux Command Line Basics • Linux Text Editors • Basic Bash Shell Scripting • Github Basics • Python Basics • Python Functions and Packages • Numpy • Plotting in Python

• Learning ROS

Module 6: Flight Geometry • 2D Coordinates and Transforms • Example • 3D Coordinates and Transforms

Module 7: Actuators & Control • Motors and Controllers • Feedback Control • Kinematics and Dynamics • Sensors

Module 8: Probability Theory and State Estimation • State Estimation • Reasoning with Bayes Rule

Module 9: Computer Vision • Introduction to Computer Vision • Image Formation • Basic Image Filtering • Edge Detection

Module 10: Control and Filters • P, PD and PID Controllers • Kalman Filters

Module 11: Visual Motion Estimation • 2D Motion Estimation • Visual Odometry

Cog*Works: Build Your Own Cognitive Assistant (2017)

For 2017, we are offering a new program called–Cog*Works: Build your own Cognitive Assistant. The program will guide students in learning and applying the foundational technologies of artificial intelligence for building cognitive assistants. This program consists of two components: (1) online course from January to May 2017, open to all interested and committed students and (2) a four-week summer program at MIT for a small group of students, July 10–August 4, 2017. During the course, the students will be trained to understand the basics of Python, Git, natural language processing, machine learning, and IBM Watson services through a series of online teaching modules. Students will build services that are both functional and fun. Students who have successfully completed the online course will be considered for participation in the summer program in which teams of students will leverage professional cognition services (e.g., IBM’s Watson services) and open-source tools in conjunction with their own machine learning tools to develop cognitive systems. The program will be divided into modules during which students will implement and explore algorithms in core areas of natural language processing and machine cognition. These capabilities will be composed to create end-to-end cognitive assistants that will compete against each other at the end of the program. By participating in the online and/or onsite portion of the program, students will develop experience in an area of computer science that is poised to play a critical role in shaping future technologies and applications across many industries.

Cog*Works Online Course Outline The BWSI Cog*Works course consists of two components: (1) an online course from January to May 2017, open to all interested and committed students and (2) a four-week summer program at MIT for a small group of students, July 10–August 4, 2017. During the online course, the students will be trained to understand the basics of Python, Git, natural language processing, machine learning, and IBM Watson services. At the completion of this online curriculum, students will be well prepared for the hands-on exercises and projects during the four-week BWSI summer program. Selected students who have successfully completed the online course will be considered for participation in the summer program.

By participating in the online and/or onsite portion of the program, students will develop experience in an area of computer science that is poised to play a critical role in shaping future technologies and applications across many industries.

Module 1: Getting Started with Python (January) • Setting up a Python Environment via Anaconda • A Brief Introduction to Jupyter Notebooks • An Informal Introduction to Python • A Quickstart Tutorial for Numerical Work in Python: NumPy

Module 2: Perspectives on Machine Learning (February) • What a Deep Neural Network Thinks About Your #selfie • The Unreasonable Effectiveness of Recurrent Neural Networks • Deep Reinforcement Learning: Pong from Pixels • The State of Computer Vision and AI • StarCraft will Become the Next Big Playground for AI

Module 3: Manage Your Work Using Git & Github (February) • What is Version Control? (Video Overview) • A Rapid, Interactive Introduction to Git • Installing Git • Git Basics • Git Branching • GitHub

Module 4: Python Revisited (March) • Control Flow Tools • Data Structures • Modules • Input and Output • Classes

Module 5: Natural Language Processing with Python: NLTK (April) • Custom NLTK Tutorial

Module 6: Simple Image Classification with Python (May) • Nearest Neighbor Classification • Linear Classification

Module 7: Learning IBM Watson© Services (June) • Learn to use IBM Watson using starter kits

o News Intelligence and Answer Retrieval o Text Chatbot and Speech Services o Visual Recongnition

• Introduction to student projects for the summer course

Online Program Application Process The Beaver Works Summer Institute is pleased to announce that we are expanding our program offerings for summer 2017. Three programs will be offered:

Mini-Grand Prix Challenge (Autonomous RACECAR) Autonomous Air Vehicle Racing (UAV Quad-Rotor) Cog*Works: Build Your Own Cognitive Assistant (Big Data)

The first step in applying to one of these summer programs is completing the BWSI Online Education Program, designed to prepare students for the technically rigorous BWSI summer programs. To participate in the online course, a student must submit the following:

1. A nomination by a school technical point of contact (POC), typically a STEM teacher or school administrator familiar with the student’s schoolwork and technical abilities.

2. The application form at the Beaver Works Summer Institute website and the required permission forms. In order to apply you must request the password from your school administrator. Once the application is approved, the student will receive an email containing a user ID and password as well as instructions on how to access the online course website.

Teachers, prior BWSI students, and teaching Assistants (TA) can also apply for online course access by using the same BWSI webpage and choosing the appropriate category from the pull-down menu.

Application for the BWSI Summer Program is separate from the online course application. The Summer Program application will be available mid-February 2017, with decisions expected in early April 2017. The selection criteria for the Summer Program include, but are not limited to,

1. Demonstrated technical ability (determined through recommendation by school official, and other supporting information such as test scores, completed coursework, and grades collected in the application).

2. Demonstrated commitment to extracurricular learning via participation and completion of the online course (participation and progress is tracked within the online course educational modules).

Students must make significant progress on the online course by March 2017 and continue their online course work into spring 2017 in order to be ready and well prepared for participation in the BWSI Summer Institute programs. Students may participate in one or more of the online courses to determine which summer course they are interested in, but note the online courses are time-intensive, and we suggest down selecting to a single course as early as possible.