Gregory Kahn

Ph.D. student, UC Berkeley EECS
Office & Mailing Address:
750 Sutardja Dai Hall
Berkeley CA 94720
Curriculum Vitae (February 2017):
I am a Ph.D. student in EECS at UC Berkeley advised by Professor Pieter Abbeel and Professor Sergey Levine in the Berkeley Artificial Intelligence Research (BAIR) Lab.

My main research goal is to develop algorithms that enable robots to operate in the real world. I am currently working on deep reinforcement learning. In the past, I have worked on trajectory optimization, planning under uncertainty, manipulation, and surgical robotics.


Neural Network Dynamics for Model-Based Deep Reinforcement Learning with Model-Free Fine-Tuning
Anusha Nagabandi, Gregory Kahn, Ronald S. Fearing, Sergey Levine
arXiv:1702.01182 [PDF][Video]

We demonstrate that medium-sized neural network models can be combined with MPC to achieve excellent sample complexity in a model-based RL algorithm, producing stable and plausible gaits to accomplish various complex locomotion tasks. We also propose using deep neural network dynamics models to initialize a model-free learner. We empirically demonstrate that this resulting hybrid algorithm can drastically accelerate model-free learning on several MuJoCo locomotion tasks.

Uncertainty-Aware Reinforcement Learning for Collision Avoidance
Gregory Kahn, Adam Villaflor, Vitchyr Pong, Pieter Abbeel, Sergey Levine
arXiv:1702.01182 [PDF][Video][Slides]

Practical deployment of reinforcement learning methods must contend with the fact that the training process itself can be unsafe for the robot. In this paper, we consider the specific case of a mobile robot learning to navigate an a priori unknown environment while avoiding collisions. We present an uncertainty-aware model-based learning algorithm that estimates the probability of collision together with a statistical estimate of uncertainty. We evaluate our method on a simulated and real-world quadrotor, and a real-world RC car.


PLATO: Policy Learning using Adaptive Trajectory Optimization
Gregory Kahn, Tianhao Zhang, Sergey Levine, Pieter Abbeel
ICRA 2017. [PDF][Video][Slides]

We propose PLATO, an algorithm that trains complex neural network policies using an adaptive variant of model-predictive control (MPC) to generate the supervision. We prove that our adaptive MPC teacher produces supervision that leads to good long-horizon performance of the resulting policy, and empirically demonstrate that MPC can avoid dangerous on-policy actions in unexpected situations during training.

   Occlusion-Aware Multi-Robot 3D Tracking
Karol Hausman, Gregory Kahn, Sachin Patil, Joerg Mueller, Ken Goldberg, Pieter Abbeel, Gaurav Sukhatme
IROS 2016.

We introduce an optimization-based control approach that enables a team of robots to cooperatively track a target using onboard sensing. The robots are required to estimate their own positions as well as tracking the target by reasoning about occlusions. We evaluate our approach in a number of experiments in which we simulate a team of quadrotor robots flying in three-dimensional space to track a moving target on the ground.

Learning Deep Control Policies for Autonomous Aerial Vehicles with MPC-Guided Policy Search
Tianhao Zhang, Gregory Kahn, Sergey Levine, Pieter Abbeel
ICRA 2016. [PDF][Video]

This paper presents a method for training neural network policies for autonomous aerial vehicles using model-predictive control (MPC) and guided policy search. A major challenge in applying reinforcement learning to aerial vehicles is the possibility of critical failure during training. To that end, MPC is used to guide off-policy learning with guided policy search. The final neural network policy provides runtime efficiency and generalization, and removes the need for explicit state estimation at test time by using raw sensor inputs.

Information-Theoretic Planning with Trajectory Optimization for Dense 3D Mapping
Benjamin Charrow, Gregory Kahn, Sachin Patil, Sikang Liu, Ken Goldberg, Pieter Abbeel, Nathan Michael, Vijay Kumar
RSS 2015. [PDF]

We propose an information-theoretic planning approach that enables mobile robots to autonomously construct dense 3D maps using a two stage approach. First, we generate a candidate set of trajectories using a combination of global planning and generation of local motion primitives. Second, we employ a gradient-based optimization to locally refine the Cauchy-Schwarz quadratic mutual information (CSQMI) objective. We evaluated our approach through a series of real-world experiments with a ground robot and simulations with an aerial robot.

Active Exploration using Trajectory Optimization for Robotic Grasping in the Presence of Occlusions
Gregory Kahn, Peter Sujan, Sachin Patil, Shaunak D. Bopardikar, Julian Ryde, Ken Goldberg, Pieter Abbeel
ICRA 2015. [PDF][Video]

We consider the task of actively exploring unstructured environments to facilitate robotic grasping of occluded objects. The objective is to plan the motion of hte sensor in order to search for feasible grasph handles that lie within occluded regions of the map. We evaluated our approach by actively exploring and attempting 300 grasps with an RGB-D sensor mounted on the end effector of a PR2 robot.

Scaling up Gaussian Belief Space Planning through Covariance-Free Trajectory Optimization and Automatic Differentiation
Sachin Patil, Gregory Kahn, Michael Laskey, John Schulman, Ken Goldberg, Pieter Abbeel
WAFR 2014. [PDF]

Belief space planning provides a principled framework to compute motion plans that explicitly gather information from sensing, as necessary, to reduce uncertainty about the robot and the environment. We consider the problem of planning in Gaussian belief spaces, which are parameterized in terms of mean states and covariances describing uncertainty. Our experiments suggest that our method can solve planning problems in 100 dimensional state spaces and obtain computational speedups of 400x over related trajectory optimization methods.

Autonomous Multilateral Debridement with the Raven Surgical Robot
Ben Kehoe, Gregory Kahn, Jeffrey Mahler, Jonathan Kim, Alex Lee, Anna Lee, Keisuke Nakagawa, Sachin Patil, W. Douglas Boyd, Pieter Abbeel, Ken Goldberg
ICRA 2014. [PDF][Video]

We present an implemented automated surgical debridement system that uses the Raven, an open-architecture surgical robot with two cable-driven 7 DOF arms. Our system combines stereo vision for 3D perception, trajopt, an optimization-based motion planner, and model predictive control (MPC).

Research Support

National Science Foundation Graduate Research Fellowship, 2016-present