Ashwin Balakrishna

I am a 3rd year PhD student at the AUTOLAB in UC Berkeley in Computer Science, with a focus in Artificial Intelligence and Robotics. I work on algorithms for scalable and safe imitation learning, reinforcement learning, and control for robotic systems and am excited about applications in robotic manipulation, including grasping, pushing, and manipulating deformable objects such as rope and cloth. I am advised by Ken Goldberg and funded by an NSF GRFP. I did my bachelors at Caltech in Electrical Engineering, where I worked on a number of applications of machine learning and signal processing to scientific problems, primarily in biomedical devices.

Email  /  CV  /  Google Scholar  

profile photo
Research

I'm interested in algorithms for imitation learning (IL), reinforcement learning (RL), and control which can be reliably deployed on robotic systems. My research is primarily focused on exploring synergies between ideas in IL, RL, and control theory to develop safe and reliable algorithms for robotic control. In addition, I am also excited about applications of machine learning, controls, and signal processing in a variety of contexts, including for robot grasping, deformable manipulation, biomedical devices, geophysics, and aeronautics.

Algorithms for Safe RL and Control
Recovery RL: Safe Reinforcement Learning with Learned Recovery Zones
Brijen Thananjeyan*, Ashwin Balakrishna*, Suraj Nair, Michael Luo, Krishnan Srinivasan, Minho Hwang, Joseph E. Gonzalez, Julian Ibarz, Chelsea Finn, Ken Goldberg
Preprint, 2020
PDF / Bibtex

An algorithm for safe reinforcement learning which utilizes a set of offline data to learn about constraints before policy learning and a pair of policies which seperate the often conflicting objectives of task directed exploration and constraint satisfaction to learn contact rich and visuomotor control tasks.

ABC-LMPC: Safe Sample-Based Learning MPC for Stochastic Nonlinear Dynamical Systems with Adjustable Boundary Conditions
Brijen Thananjeyan*, Ashwin Balakrishna*, Ugo Rosolia, Joseph E. Gonzalez, Aaron Ames, Ken Goldberg
Algorithmic Foundations of Robotics (WAFR), 2020
PDF / Bibtex

An MPC-based algorithm for robotic control (ABC-LMPC) with (1) performance and safety guarantees for stochastic nonlinear systems and (2) the ability to continuously explore the environment and expand the controller domain.

Safety Augmented Value Estimation from Demonstrations (SAVED): Safe Deep Model-Based RL for Sparse Cost Robotic Tasks
Brijen Thananjeyan*, Ashwin Balakrishna*, Ugo Rosolia, Felix Li, Rowan McAllister, Joseph E. Gonzalez, Sergey Levine, Francesco Borrelli, Ken Goldberg
Robotics and Automation Letters (RA-L) and International Conference on Robotics and Automation (ICRA), 2020
Website / PDF / Bibtex

A new algorithm for safe and efficient reinforcement learning (SAVED) which leverages a small set of suboptimal demonstrations and prior task successes to structure exploration. SAVED also provides a mechanism for handling state-space constraints by leveraging probabilistic estimates of system dynamics.

Efficient Online Learning for Robotics
Exploratory Grasping: Performance Bounds and Asymptotically Optimal Algorithms for Learning to Robustly Grasp an Unknown Polyhedral Object
Michael Danielczuk*, Ashwin Balakrishna*, Daniel Brown, Ken Goldberg
Preprint, 2020
PDF / Bibtex

An asymptotically optimal algorithm for exploring grasps across different stable poses of an object with unknown geometry.

Accelerating Grasp Exploration by Leveraging Learned Priors
Katherine Li*, Michael Danielczuk*, Ashwin Balakrishna*, Vishal Satish, Ken Goldberg
Conference on Automation Science and Engineering (CASE), 2020
PDF / Bibtex

An algorithm for leveraging priors from general purpose grasping systems to accelerate online grasp exploration on novel, difficult to grasp objects.

On-Policy Robot Imitation Learning from a Converging Supervisor
Ashwin Balakrishna*, Brijen Thananjeyan*, Jonathan Lee, Felix Li, Arsh Zahed, Joseph E. Gonzalez, Ken Goldberg
Conference on Robot Learning (CoRL) - Oral Presentation, 2020
PDF / Bibtex

A new formulation of imitiation learning from a non-stationary supervisor, associated theoretical analysis, and a practical algorithm to apply this formulation to develeop an RL algorithm which combines the sample efficiency of model-based RL and the fast policy evaluation enabled by model-free policies.

Visuomotor Control with Simulated Supervision
Orienting Novel 3D Objects Using Self-Supervised Learning of Rotation Transforms
Shivin Devgon, Jeffrey Ichnowski, Ashwin Balakrishna, Harry Zhang, Ken Goldberg
Conference on Automation Science and Engineering (CASE), 2020
PDF / Bibtex

A self-supervised algorithm which learns to orient unseen objects with unknown geometry given only a depth image observation of the desired orientation.

Learning to Smooth and Fold Real Fabric Using Dense Object Descriptors Trained on Synthetic Color Images
Aditya Ganapathi, Priya Sundaresan, Brijen Thananjeyan, Ashwin Balakrishna, Daniel Seita, Jennifer Grannen, Minho Hwang, Ryan Hoque, Joseph E. Gonzalez, Nawid Jamali, Katsu Yamane, Soshi Iba, Ken Goldberg
Preprint, 2020
Website / PDF / Bibtex

A general method for multi-task fabric manipulation using learned visual correspondences which can be applied across different robots to manipulate fabrics of varying shapes and colors.

VisuoSpatial Foresight (VSF) for Multi-Step, Multi-Task Fabric Manipulation
Ryan Hoque*, Daniel Seita*, Ashwin Balakrishna, Aditya Ganapathi, Ajay Tanwani, Nawid Jamali, Katsu Yamane, Soshi Iba, Ken Goldberg
Robotics Science and Systems (RSS), 2020
Website / PDF / Bibtex

A method for multi task fabric manipulation by leveraging recent advances in video prediction and depth sensing.

Deep Imitation Learning of Sequential Fabric Smoothing from an Algorithmic Supervisor
Daniel Seita, Aditya Ganapathi, Ryan Hoque, Minho Hwang, Edward Cen, Ajay Kumar Tanwani, Ashwin Balakrishna, Brijen Thananjeyan, Jeffrey Ichnowski, Nawid Jamali, Katsu Yamane, Soshi Iba, John F. Canny, Ken Goldberg
International Conference on Intelligent Robots and Systems (IROS), 2020
Website / PDF / Bibtex

A new fabric simulator for learning fabric smoothing policies and learned policies which successfully smooth fabric in simulation and transfer to physical robotic systems.

Learning Interpretable and Transferable Rope Manipulation Policies Using Depth Sensing and Dense Object Descriptors
Priya Sundaresan, Jennifer Grannen, Brijen Thananjeyan, Ashwin Balakrishna, Michael Laskey, Kevin Stone, Joseph E. Gonzalez, Ken Goldberg
International Conference on Robotics and Automation (ICRA), 2020
Website / PDF / Bibtex

An algorithm for learning visual correspondences for highly deformable objects and an associated controller which is used to manipulate rope into a variety of different arrangements either by learning from demonstrations or by designing interpretable geometric policies on top of the learned visual representation.

Mechanical Search: Multi-Step Retrieval of a Target Object Occluded by Clutter
Michael Danielczuk*, Andrey Kurenkov*, Ashwin Balakrishna, Matthew Matl, David Wang, Roberto Martin-Martin, Animesh Garg, Silvio Savarase, Ken Goldberg
International Conference on Robotics and Automation (ICRA), 2019
Website / PDF / Bibtex

Formulation and algorithms for the problem of efficiently identifying and retrieving a specific object in a cluttered environment.

Biomedical Signal Processing
Fabry-PĂ©rot optical sensor and portable detector for monitoring high-resolution ocular hemodynamics
Jeong Oen Lee, Vinayak Narasimhan, Ashwin Balakrishna, Marcus R. Smith Juan Du, David Stretavan, Hyuck Choo
Photonics Technology Letters, 2019
PDF / Bibtex

High resolution measurement of both intraocular pressure and ocular pulsation profiles using an implmantable micro-optical sensor and portable optical detector.

Machine Learning Methods for Rapid, Real-Time Pressure Readout from an Optics-Based Intraocular Pressure Sensor
Ashwin Balakrishna, Jeong Oen Lee, Hyuck Choo
Preprint, 2018
PDF / Bibtex

An evaluation of different machine learning algorithms for intraocular pressure measurement.

A microscale optical implant for continuous in vivo monitoring of intraocular pressure
Jeong Oen Lee, Haeri Park, Juan Du, Ashwin Balakrishna, Oliver Chen, David Stretavan, Hyuck Choo
Nature: Microsystems and Nanoengineering, 2017
PDF / Bibtex

A new microscale implantable sensor and associated algorithms for accurate and convenient measurement of intraocular pressure.

Novel positioning sensor with real-time feedback for improved postoperative positioning: pilot study in control subjects
Frank Brodie, David Ramirez*, Sundar Pandian*, Kelly Woo, Ashwin Balakrishna, Eugene De Juan, Hyuck Choo, Robert H Grubbs
Clinical Opthalmology, 2017
PDF / Bibtex

A new wearable, wireless sensor to aid postoperative recovery from retinal detachment surgery.

Validation of sensor for postoperative positioning with intraocular gas
Frank Brodie, Kelly Woo, Ashwin Balakrishna, Hyuck Choo, Robert H Grubbs
Clinical Opthalmology, 2016
PDF / Bibtex

A simple, wearable electromechanical sensor to aid postoperative recovery from retinal detachment surgery.

A Neural Network Approach to Monitor Intraocular Pressure for Glaucoma Diagnosis
Ashwin Balakrishna, Oliver Chen, Jeong Oen Lee, Hyuck Choo
Progress In Electromagnetics Research Symposium (PIERS), 2016
PDF / Bibtex

A neural network based algorithm to efficiently extract accurate intraocular pressure measurements from reflection spectra from an optics-based intraocular pressure sensor.

In vivo Intraocular Pressure Monitoring using Implantable Optomechanical Sensor
Jeong Oen Lee, Haero Park, Juan Du, Vinayak Narasimhan, Ashwin Balakrishna, Oliver Chen, David Stretavan, Hyuck Choo
Interenational Symposium on Optomechatronic Technology (ISOT), 2016
PDF / Bibtex

Evaluation of a new optics-based intraocular pressure sensor in live rabbits.

Efficient Power Generation from Vocal Folds Vibrations for Medical Electronic Implants
Hyunjun Cho, Ashwin Balakrishna, Yuan Ma, Jeong Oen Lee, Hyuck Choo
International Conference on Micro-Electro-Mechanical Systems (MEMS), 2016
PDF / Bibtex

A piezoelectric based device to harvest power from human vocal cord vibrations.

Miscellaneous
Automating Planar Object Singulation by Linear Pushing with Single-point and Multi-point Contacts
Zisu Dong, Sanjay Krishnan, Sona Dolasia, Ashwin Balakrishna, Michael Danielczuk, Ken Goldberg
Conference on Automation Science and Engineering (CASE), 2019
PDF / Bibtex

An efficient geometric algorithm (ClusterPush) for singulating a set of clustered planar objects.

Reliable Real-time Seismic Signal/Noise Discrimination with Machine Learning
Men-Andrin Meier, Zachary E. Ross, Anshul Ramachandran Ashwin Balakrishna, Suraj Nair, Peter Kundzicz, Zefeng Li, Jennifer Andrews, Egill Hauksson, Yisong Yue
Journal of Geophysical Research: Solid Earth, 2018
PDF / Bibtex

Algorithms for rapid and reliable earthquake detection for earthquake early warning systems.

Predicting Electric Vehicle Charging Station Usage: Using Machine Learning to Estimate Individual Station Statistics from Physical Configurations of Charging Station Networks
Anshul Ramachandran, Ashwin Balakrishna, Peter Kundzicz, Anirudh Neti
Preprint, 2018
PDF / Bibtex

Algorithms for predicting electrical vehicle power usage for different charging network designs.

Optimal Control Strategies for Trajectory Optimization with Applications to Continuous Solar Flight
Ashwin Balakrishna
INFORMS Annual Meeting, High School Mathematical Science Journal, Intel Science Talent Search Semifinalist, 2014
PDF / Bibtex

Mathematical model and algorithm for controlling continuously flying solar aircraft.


Website template from Jon Barron