For the most up to date list, you can also check Rob's google scholar profile.
AutOTranS: an Autonomous Open World Transportation System
Zapata-Impata, B., Shah, V., Singh, H., Platt, R. Project page Summary: We propose an approach to autonomous transport of a wide class of novel objects in unstructured settings outdoors. The user identifies what needs to be moved (e.g. move the trash) and where it needs to go (e.g. in the trash bin). The system does the rest autonomously! I could use one of these at my house!
Learning Manipulation Skills Via Hierarchical Spatial Attention
Gualtieri, M., Platt, R. Summary: On of the key challenges in learning model free policies for pick and place policies is the high dimensional action space. One approach to the problem is to set action variable values hierarchically -- by learning a policy for sequentially constraining the action selection until a final action selection is made. This paper explores the approach in the context of 6-DOF reach action selection.
Online abstraction with MDP homomorphisms for Deep Learning
Biza, O., Platt, R.
AAMAS 2019 Summary: MDP abstraction is an attractive idea for simplifying policy learning. Instead of attempting to find control policy in the original low-level state and action space, the idea is to learn a compact MDP representation of the original problem that can easily be solved and then to project that compact solution back to the original problem. The key challenge here is finding the abstraction. This paper proposes an approach to finding a particular type of abstraction, called an MDP homomorphism. We focus on abstraction scenarios relevant to robotics: those with large and continuous state action spaces. As far as we know, this is the first approach to MDP homomorphism finding that works outside of the finite MDP setting.
Learning Multi-Level Hierarchies with Hindsight
Levy, A., Konidaris, G., Platt, R., Saenko, K.
ICLR 2019 Summary: Reinforcement Learning (RL) algorithms can suffer from poor sample efficiency when rewards are delayed and sparse. We introduce a solution that enables agents to learn temporally extended actions at multiple levels of abstraction in a sample efficient and automated fashion. Our approach combines universal value functions and hindsight learning, allowing agents to learn policies belonging to different time scales in parallel. We show that our method significantly accelerates learning in a variety of discrete and continuous tasks.
Towards Assistive Robotic Pick and Place in Open World Environments
Wang, D., Kohler, C., ten Pas, A., Wilkinson, A., Liu, M., Yanco, H., Platt, R.
ISRR 2019 Project page Summary: This paper describes an assistive robotic mobility scooter that can assist users with the activities of daily life (ADLs). The user points at a novel object to be picked up using a laser pointer and the system autonomously performs the grasp. It can also handle placing. This type of system could make sense for many mobility scooter users who have difficulty reaching down to pick things up.
Deictic Image Maps: An Abstraction For Learning Pose Invariant Manipulation Policies
Platt, R., Kohler, C., Gualtieri, M.
AAAI 2019 Slides Summary: In applications of deep reinforcement learning to robotics, we often want to learn pose invariant policies: policies that are invariant to changes in the position and orientation of objects in the world. This paper proposes a state and action abstraction that encodes such invariance. As a result, we can learn manipulation policies that generalize better to the unstructured world than would otherwise be posible.
Learning 6-DoF Grasping and Pick-Place Using Attention Focus
Gualtieri, M., Platt, R.
CoRL 2018 Summary:
We address a class of manipulation problems where the robot perceives the scene with a depth sensor and can move its end effector in a space with six degrees of freedom -- 3D position and orientation. We develop a specific set of constraints called hierarchical SE(3) sampling (HSE3S) which causes the robot to learn a sequence of gazes to focus attention on the task-relevant parts of the scene. We demonstrate the effectiveness of our approach on three challenging pick-place tasks both in simulation and on a real robot.
Adapting control policies from simulation to reality using a pairwise loss
Viereck, U., Saenko, K., Platt, R.
ISER 2018. Summary:
This paper proposes an approach to domain transfer based on a pairwise loss function that helps transfer control policies learned in simulation onto a real robot. Experimental results demonstrate that proposed method consistently outperforms baseline methods that train only in simulation or that combine real and simulated data in a naive way.
Pick and Place Without Geometric Object Models
Gualtieri, M., Ten Pas, A., Platt, R.
ICRA 2018. Summary: This paper proposes a new formulation of robotic pick and place as a deep RL problem where the actions are grasp and place poses for the robot's hand, and the state is encoded with the observed geometry local to a selected grasp. This framework enables us to learn to perform pick and place tasks with novel objects in clutter.
Grasp Pose Detection in Point Clouds
Gualtieri, M., Ten Pas, A., Saenko, K., Platt, R.
Int'l Journal of Robotics Research (IJRR), Vol 36, No 13-14, pp 1455--1473, 2017 Code available here. Summary: This is the journal length paper describing GPD 1.0.0, our latest grasp detection work. Using this method, our robot is able to grasp novel objects with a 93% grasp success rate.
Open World Assistive Grasping Using Laser Selection
Gualtieri, M., Kuczynski, J., Shultz, A., Ten Pas, A., Yanco, H., Platt, R.
ICRA 2017. Summary: This paper describes our work integrating our grasp detection software, GPD 1.0.0, with an assistive robotic scooter that can grasp novel objects identified using a laser pointer.
Platt, R., Ihrke, C., Bridgwater, L., Linn, M., Diftler, M., Abdallah, M., Askew, S., Permenter, F., A miniature load cell suitable for mounting on the phalanges of human-sized robot fingers, IEEE Int'l Conf. on Robotics and Automation, Shanghai, May, 2011
Diftler, M., Mehling, J., Abdallah, M., Radford, N., Bridgwater, L., Sanders, A., Askew, S., Linn, D., Yamokoski, J., Permenter, F., Hargrave, B., Platt, R., Savely, R., Ambrose, R., Robonaut 2: The First Humanoid Robot in Space, IEEE Int'l Conf. on Robotics and Automation, Shanghai, May, 2011
Platt, R., Abdallah, M., Wampler, C., Multiple-priority impedance control, IEEE Int'l Conf. on Robotics and Automation, Shanghai, May, 2011
Robert Platt, Mars Chu, Myron Diftler, Toby Martin, Michael Valvo,
A Miniature Force Sensor for Prosthetic Hands, Workshop on
Robotic Systems for Rehabilitation, Exoskeleton, and Prosthetics,
Robotics: Science and Systems, University of Pennsylvania,
Philadelphia, PA, August 18, 2006.
William Bluethmann, Robert Ambrose, Myron Diftler, Eric Huber, Andy
Fagg, Michael Rosenstein, Robert Platt, Roderic Grupen, Cynthia
Breazeal, Andrew Brooks, Andrea Lockerd, R. Alan Peters II, O. Chad
Jenkins, Maja Mataric, Magdalena Bugajska
Building an Autonomous Humanoid Tool User,
Proceedings of the 2004 IEEE International
Conference on Humanoid Robots, Los Angeles, CA, USA November 2004
A. Fagg, M. Rosenstein, R. Platt and R. Grupen,
Extracting User Intent in Mixed Initiative Teleoperator
Control, AIAA-2004-6309 AIAA 1st Intelligent Systems Technical
Conference, Chicago, Illinois, Sep. 20-22, 2004
T. Martin, M. Diftler and R. Ambrose, R. Platt, M. Butzer,
Tactile Sensors for the NASA/DARPA Robonaut, AIAA 1st
Intelligent Systems Technical Conference, Chicago, Illinois,
Sep. 20-22, 2004