Robert Platt Jr.

 

Current Projects and Research Directions


Affordance-based perception for grasping novel objects

Robust robot grasping and manipulation depends on accurately localizing appropriate grasp geometries. Rather than attempt to localize all relevant objects in the scene, we have developed an approach to localizing handle-like geometries on novel objects that can be grasped by encircling it with the thumb and fingers of the robot hand. For example, rather than localizing a coffee mug and creating a plan to grasp it, our system localizes and grasp the mug handle independently of object identity. Our method is based on modeling graspable geometries as hollow cylindrical shells. We fit an implicit quadratic function (in three variables) to neighborhoods of a point cloud using a least squares algebraic fit with Taubin normalization. This enables us to identify a set of candidate cylindrical hollow shells. Shells that to not contain empty space between the inner and outer cylinders that would allow the robot hand to encircle the object are discarded.

New: You can download a ROS package implementing our handle finder. It's easy to use: all you need is an Asus Xtion Pro and to install ROS, PCL, and our package.

Robot grasp
Grasp affordances localized by our algorithm
Novel objects
All handles found in the scene are shown in cyan. Red shows the selected grasp target.

Ten Pas, A., Platt, R., Localizing Grasp Affordances in 3-D Points Clouds Using Taubin Quadric Fitting, arXiv:1311.3192 [cs.RO], First submitted: November, 2013


Tactile SLAM

A key challenge in robot manipulation is localizing an object (or part of an object) while the robot is holding it. One way to accomplish this is to use tactile sensing. In our work, a tactile map of an object is created that describes what an object "feels like" to the robot (it is a model of tactile sensations as a function of contact configuration). When an object is held during manipulation, it can be localized in the hand by matching to this map. We expect this technology will be very useful in the context of automated insertions of small or flexible parts such as the USB cable insertion shown at right.

Tactile map of a coin formed using the Gelsight tactile sensor.
Tactile sensing can make this kind of insertion more robust.

Platt, R., Permenter, F., Pfeiffer, J., Using Bayesian filtering to localize flexible materials during manipulation, IEEE Transactions on Robotics, Special issue on a robotic sense of touch. Vol 27, No 3, June 2011

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


Planning Under Uncertainty in Manipulation

One of the main research challenges in robot manipulation is developing reliable grasping capabilities. It is difficult to guarantee accurate perception of the objects to be grasped or manipulated. In these contexts, it is insufficient just to identify the most likely state of the world; the system must be aware of what it knows and what is not known. Furthermore, when the state of the world is uncertain, the system must be capable of acting in order to gain information. We are developing new approaches to planning under uncertainty that can do this kind of reasoning in robot manipulation scenarios. For example, we are using planning under uncertainty to reason about how to position a camera or range sensor in order to more accurately localize relevant objects to be grasped. We are also using planning under uncertainty to reason about how to push objects in order to better localize them.

The planner pushes the left box away from the right so that it will better perceive both boxes.

Platt, R., Kaelbling, L., Lozano-Perez, T., Tedrake, R. Non-Gaussian Belief Space Planning: Correctness and Complexity, IEEE Int'l Conf. on Robotics and Automation, 2012. (The final version of the paper posted here fixes some errors that were present in the proofs in the submitted version.)

Platt, R., Kaelbling, L., Lozano-Perez, T., Tedrake, R.Efficient planning in non-Gaussian belief spaces and its application to robot grasping, Proceedings of the International Symposium on Robotics Research, 2011. (Extended version available in CSAIL Tech Report MIT-CSAIL-TR-2011-039

Platt, R., Tedrake, R., Kaelbling, L., Lozano-Perez, T., Belief space planning assuming maximum likelihood observations, Proceedings of Robotics: Science and Systems 2010 (RSS), Zaragosa, Spain, June 27, 2010.


The DARPA Robotics Challenge

Our group has recently been involved with two different DARPA Robotics Challenge teams: NASA-JSC Team Valkyrie and Team TRACLabs. We are helping these two teams develop superior grasping and manipulation related capabilities for their respective robots. Update: team TRACLabs recently placed 6th in the DARPA Robotics Challenge Trials. We are very excited!

Ryde, J., Dhiman, V., Platt, R., Voxel Planes: Rapid Visualization and Meshification of Point Cloud Ensembles, IEEE Int'l Conf. on Intelligent Robotics Systems (IROS), 2013.


LQR-RRT*

The robot motion planning problem is very important and arises in several types of robotics problems including manipulation planning, mobile robot motion planning, and aerial vehicle motion planning. RRT* is a recently-developed motion planning algorithm extends the Rapidly Exploring Random Tree (RRT). It is significant because, unlike RRT, it is guaranteed (with probability one) to converge to an optimal motion planning solution. Recently, we have extended the RRT* algorithm to handle arbitrary linear quadratic differentially constrained systems by incorporating ideas from linear quadratic regulation (LQR).

The path found by our algorithm for a double-integrator system.

Goretkin, G., Perez, A., Platt, R., Konidaris, G., Optimal Sampling-Based Planning for Linear-Quadratic Kinodynamic Systems, IEEE Int'l Conf. on Robotics and Automation (ICRA), 2013.

Perez, A., Platt, R., Konidaris, G., Kaelbling, L., Lozano-Perez, T. LQR-RRT^*: Optimal Sampling-Based Motion Planning with Automatically Derived Extension Heuristics, IEEE Int'l Conf. on Robotics and Automation, 2012. (This paper has been updated since our original submission.)