Michael C. Koval

Pose estimation for contact manipulation with manifold particle filters

We investigate the problem of estimating the state of an object during manipulation. Contact sensors provide valuable information about the object state during actions which involve persistent contact, e.g. pushing. However, contact sensing is very discriminative by nature, and therefore the set of object states that contact a sensor constitutes a lower-dimensional manifold in the state space of the object. This causes stochastic state estimation methods, such as particle filters, to perform poorly when contact sensors are used. We propose a new algorithm, the manifold particle filter, which uses dual particles directly sampled from the contact manifold to avoid this problem. The algorithm adapts to the probability of contact by dynamically changing the number of dual particles sampled from the manifold. We compare our algorithm to the conventional particle filter through extensive experiments and we show that our algorithm is both faster and better at estimating the state. Unlike the conventional particle filter, our algorithms performance improves with increasing sensor accuracy and the filters update rate. We implement the algorithm on a real robot using commercially available tactile sensors to track the pose of a pushed object.

Pre- and post-contact policy decomposition for planar contact manipulation under uncertainty

We consider the problem of using real-time feedback from contact sensors to create closed-loop pushing actions. To do so, we formulate the problem as a partially observable Markov decision process (POMDP) with a transition model based on a physics simulator and a reward function that drives the robot towards a successful grasp. We demonstrate that it is intractable to solve the full POMDP with traditional techniques and introduce a novel decomposition of the policy into pre- and post-contact stages to reduce the computational complexity. Our method uses an offline point-based solver on a variable-resolution discretization of the state space to solve for a post-contact policy as a pre-computation step. Then, at runtime, we use an A* search to compute a pre-contact trajectory. We prove that the value of the resulting policy is within a bound of the value of the optimal policy and give intuition about when it performs well. Additionally, we show the policy produced by our algorithm achieves a successful grasp more quickly and with higher probability than a baseline QMDP policy on two different objects in simulation. Finally, we validate our simulation results on a real robot using commercially available tactile sensors.

Object search by manipulation

We investigate the problem of a robot searching for an object. This requires reasoning about both perception and manipulation: certain objects are moved because the target may be hidden behind them and others are moved because they block the manipulators access to other objects. We contribute a formulation of the object search by manipulation problem using visibility and accessibility relations between objects. We also propose a greedy algorithm and show that it is optimal under certain conditions. We propose a second algorithm which is optimal under all conditions. This algorithm takes advantage of the structure of the visibility and accessibility relations between objects to quickly generate optimal plans. Finally, we demonstrate an implementation of both algorithms on a real robot using a real object detection system.

Pose estimation for planar contact manipulation with manifold particle filters

We investigate the problem of using contact sensors to estimate the pose of an object during planar pushing by a fixed-shape hand. Contact sensors are unique because they inherently discriminate between “contact” and “no-contact” configurations. As a result, the set of object configurations that activates a sensor constitutes a lower-dimensional contact manifold in the configuration space of the object. This causes conventional state estimation methods, such as the particle filter, to perform poorly during periods of contact due to particle starvation. In this paper, we introduce the manifold particle filter as a principled way of solving the state estimation problem when the state moves between multiple manifolds of different dimensionality. The manifold particle filter avoids particle starvation during contact by adaptively sampling particles that reside on the contact manifold from the dual proposal distribution. We describe three techniques, one analytical and two sample-based, of sampling from the dual proposal distribution and compare their relative strengths and weaknesses. We present simulation results that show that all three techniques outperform the conventional particle filter in both speed and accuracy. In addition, we implement the manifold particle filter on a real robot and show that it successfully tracks the pose of a pushed object using commercially available tactile sensors.

HERB's Sure Thing: A rapid drama system for rehearsing and performing live robot theater

In Spring 2014, the Personal Robotics Lab at CMU collaborated with the School of Drama to develop, produce and stage a live theatrical performance at the Purnell Center for the Arts in Pittsburgh. This paper describes some of our unique experiences collaborating with drama faculty, the director and the actor. We highlight the challenges arising from theatrical performance and specifically describe some of the technical tools we developed: a bidirectional Blender interface for robot animation, an interactive system for manipulating speech prosody, and a conductors console for online improvisation and control during rehearsal and performance. It also explores some of the remaining challenges to our goal of developing algorithms and open-source tools that can enable any roboticist in the world to create their own dramatic performance.

Manifold Representations for State Estimation in Contact Manipulation

We investigate the problem of using contact sensors to estimate the configuration of an object during manipulation. Contact sensing is very discriminative by nature and, therefore, the set of object configurations that activate a sensor constitutes a lower-dimensional manifold in the configuration space of the object. This causes conventional state estimation methods, such as particle filters, to perform poorly during periods of contact. The manifold particle filter addresses this problem by sampling particles directly from the contact manifold. When it exists, we can sample these particles from an analytic representation of the contact manifold. We present two alternative sample-based contact manifold representations that make no assumptions about the object-hand geometry: rejection sampling and trajectory rollouts. We discuss theoretical considerations behind these three representations and compare their performance in a suite of simulation experiments. We show that all three representations enable the manifold particle filter to outperform the conventional particle filter. Additionally, we show that the trajectory rollout representation performs similarly to the analytic method despite the rollout method’s relative simplicity.

Robust trajectory selection for rearrangement planning as a multi-armed bandit problem

We present an algorithm for generating open-loop trajectories that solve the problem of rearrangement planning under uncertainty. We frame this as a selection problem where the goal is to choose the most robust trajectory from a finite set of candidates. We generate each candidate using a kinodynamic state space planner and evaluate it using noisy rollouts. Our key insight is we can formalize the selection problem as the “best arm” variant of the multi-armed bandit problem. We use the successive rejects algorithm to efficiently allocate rollouts between candidate trajectories given a rollout budget. We show that the successive rejects algorithm identifies the best candidate using fewer rollouts than a baseline algorithm in simulation. We also show that selecting a good candidate increases the likelihood of successful execution on a real robot.

Hybrid control trajectory optimization under uncertainty

Trajectory optimization is a fundamental problem in robotics. While optimization of continuous control trajectories is well developed, many applications require both discrete and continuous, i.e. hybrid controls. Finding an optimal sequence of hybrid controls is challenging due to the exponential explosion of discrete control combinations. Our method, based on Differential Dynamic Programming (DDP), circumvents this problem by incorporating discrete actions inside DDP: we first optimize continuous mixtures of discrete actions, and, subsequently force the mixtures into fully discrete actions. Moreover, we show how our approach can be extended to partially observable Markov decision processes (POMDPs) for trajectory planning under uncertainty. We validate the approach in a car driving problem where the robot has to switch discrete gears and in a box pushing application where the robot can switch the side of the box to push. The pose and the friction parameters of the pushed box are initially unknown and only indirectly observable.

The manifold particle filter for state estimation on high-dimensional implicit manifolds

We estimate the state of a noisy robot arm and underactuated hand using an implicit Manifold Particle Filter (MPF) informed by contact sensors. As the robot touches the world, its state space collapses to a contact manifold that we represent implicitly using a signed distance field. This allows us to extend the MPF to higher (six or more) dimensional state spaces. Earlier work, which explicitly represents the contact manifold, was only capable of scaling to three dimensions. Through a series of experiments, we show that the implicit MPF converges faster and is more accurate than a conventional particle filter during periods of persistent contact. We present three methods of drawing samples from an implicit contact manifold, and compare them in experiments.

A System for Multi-step Mobile Manipulation: Architecture, Algorithms, and Experiments

Household manipulation presents a challenge to robots because it requires perceiving a variety of objects, planning multi-step motions, and recovering from failure. This paper presents practical techniques that improve performance in these areas by considering the complete system in the context of this specific domain. We validate these techniques on a table-clearing task that involves loading objects into a tray and transporting it. The results show that these techniques improve success rate and task completion time by incorporating expected real-world performance into the system design.