Alex LaGrassa

About

I am a robotics PhD student in the Robotics Institute at Carnegie Mellon working in the Intelligent Autonomous Manipulation lab with Oliver Kroemer. My motivation is to improve autonomous robotic manipulation so robots can address changing needs in domains such as health care, service, and agriculture. I believe that the most progress will be made by effectively combining prior knowledge with information learned from data.

Education

Carnegie Mellon University - PhD (Student)

  • Robotics Institute

  • Aug 2019 - Ongoing

  • Advisor: Oliver Kroemer

  • Combining model-based planning and learning to complete contact-rich manipulation tasks.

Massachusetts Institute of Technology - Masters of Engineering

  • Computer Science and Artificial Intelligence Laboratory (CSAIL)

  • Jan 2018 - Aug 2019

  • Advisor: Leslie Kaelbling

  • Incorporating learned skills into task and motion planning by learning both the preconditions, and the policy. Investigating how to apply various reinforcement learning techniques to benchmark manipulation tasks.

Massachusetts Institute of Technology - Bachelor of Science

  • Computer Science (Course 6)

  • Aug 2014 - Jun 2018

  • Advisor: Leslie Kaelbling

  • Resolving references through planning in robotic mobile manipulation domains

Research Experience

Microsoft Research

Summer Research Intern

  • Used tactile feedback from a BioTac sensor on a Shadow Hand to train policies for manipulation for better generalization

  • Implemented BioTac sensor simulation in MuJoCo

  • Combining multiple manipulation policies

CMU Robotics Institute

Graduate Researcher

  • Planning using multiple models, learned and pre-defined

  • Contact-rich manipulation by combining force and position controllers

  • Exploring techniques to make model-based planning in manipulation domains more reliable by learning local policies where the model fails

MIT Computer Science and Artificial Intelligence Laboratory

Undergraduate and Graduate Researcher

  • Incorporating reference resolution into planning

  • Using machine learning to train primitive motor skills such as stirring, erasing, scooping, pushing, and pouring

  • Designed an encoding technique to predict success of candidate force sensor traces for generating trajectories with dynamic movement primitives.

MIT Lincoln Laboratory

Summer Undergraduate Researcher

  • Clustering algorithms

  • Radar automation scripting

Work Experience

Microsoft - Software Engineering Intern

  • Summer 2016, Summer 2017

  • Customer support ticket clustering

  • Anomaly detection using telemetry for server error prediction

  • Improving geocoding precision and recall for problematic queries

Teaching

Elements of Software Construction

  • Full time teaching assistant

  • Mentored ten project groups for a final project designing a large software application

Foundations of Information Policy

  • Mentored two project groups in a semester long research project on Internet of Things (IoT) policy.

Skills

Programming

  • Python

  • Java

  • MATLAB

  • C++

  • ML frameworks: Tensorflow, PyTorch, Keras

  • Simulators: PyBullet, MuJoCo, Gazebo, Drake

  • Planning: MoveIt, pybullet-planning, pddlstream, HPN,

  • ROS


Algorithms

  • Machine Learning

  • Reinforcement Learning

  • Optimization

Robotics

  • Decision making under uncertainty

  • Motion planning

  • Visual odometry

  • Computer Vision

  • State estimation

  • Control

  • Manipulation