We present a new approach to motion planning using stochastic trajectory optimization framework. The relies on generating noisy trajectories explore the space around an initial (possibly infeasible) trajectory, which are then combined produced updated with lower cost. A cost function based combination of obstacle and smoothness is optimized in each iteration. No gradient information required for particular algorithm that we use so general costs derivatives may not be available (e.g....

We present a new collision and proximity library that integrates several techniques for fast accurate checking computation. Our is based on hierarchical representations designed to perform multiple queries different model representations. The set of includes discrete detection, continuous separation distance computation penetration depth estimation. input models may correspond triangulated rigid or deformable articulated models. Moreover, FCL can probabilistic between noisy point clouds are...

We present a novel robotic grasp controller that allows sensorized parallel jaw gripper to gently pick up and set down unknown objects once location has been selected. Our approach is inspired by the control scheme humans employ for such actions, which known centrally depend on tactile sensation rather than vision or proprioception. processes measurements from gripper's fingertip pressure arrays hand-mounted accelerometer in real time generate signals are designed mimic human SA-I, FA-I,...

Developing robot agnostic software frameworks involves synthesizing the disparate fields of robotic theory and engineering while simultaneously accounting for a large variability in hardware designs control paradigms. As capabilities increase, setup difficulty learning curve new users also increase. If entry barriers configuring using on robots is too high, even most powerful are useless. A growing need exists to aid getting started with, customizing, framework as necessary particular...

Learning complex motor skills for real world tasks is a hard problem in robotic manipulation that often requires painstaking manual tuning and design by human expert. In this work, we present Reinforcement based approach to acquiring new from demonstration. Our allows the robot learn fine significantly improve its success rate skill level starting possibly coarse aims incorporate task domain knowledge, where appropriate, working space consistent with constraints of specific task. addition,...

Robotic grasping in unstructured environments requires the ability to select grasps for unknown objects and execute them while dealing with uncertainty due sensor noise or calibration errors. In this work, we propose a simple but robust approach grasp selection objects, reactive adjustment deal object location shape. The method uses 3D data directly determine ranked set of scene, using heuristics based on both overall shape its local features. tactile feedback from fingertip sensors...

Robots exhibit flexible behavior largely in proportion to their degree of knowledge about the world. Such is often meticulously hand-coded for a narrow class tasks, limiting scope possible robot competencies. Thus, primary factor capabilities not physical attributes robot, but limited time and skill expert programmers. One way deal with vast number situations environments that robots face outside laboratory provide users simple methods programming do require an expert. For this reason,...

We describe an autonomous robotic system capable of navigating through office environment, opening doors along the way, and plugging itself into electrical outlets to recharge as needed. demonstrate extensive experimentation that our robot executes these tasks reliably, without requiring any modification environment. present robust detection algorithms for doors, door handles, plugs sockets, combining vision laser sensors. show how overcome unavoidable shortcoming perception by integrating...

In recent years the Robot Operating System (ROS) has become 'de facto' standard framework for robotics software development. The `ros_control` provides capability to implement and manage robot controllers with a focus on both _real-time performance_ _sharing of controllers_ in robot-agnostic way. primary motivation sepate robot-control is lack realtime-safe communication layer ROS. Furthermore, implements solutions controller-lifecycle hardware resource management as well abstractions...

Unstructured human environments present a substantial challenge to effective robotic operation. Mobile manipulation in requires dealing with novel unknown objects, cluttered workspaces, and noisy sensor data. We an approach mobile pick place such using combination of two-dimensional (2-D) three-dimensional (3-D) visual processing, tactile proprioceptive data, fast motion planning, reactive control monitoring, grasping. demonstrate our by two-arm system objects. Reactive components allow...

Heuristic searches such as A* search are highly popular means of finding least-cost plans due to their generality, strong theoretical guarantees on completeness and optimality simplicity in the implementation. In planning for robotic manipulation however, these techniques commonly thought impractical high-dimensionality problem. this paper, we present a heuristic search-based planner that does deal effectively with The achieves required efficiency following three factors: (a) its use...

Reconfigurable modular robots have the ability to use different gaits and configurations perform various tasks. A rolling gait is fastest currently implemented available for traversal over level ground shows dramatic improvements in efficiency. In this work, we analyze implement a sensor-based feedback controller achieve dynamic loop robot. The robot senses its position relative changes shape as it rolls. This such that center of gravity maintained be front contact point with ground, so...

Tactile information is valuable in determining properties of objects that are inaccessible from visual perception. In this paper, we present a tactile perception strategy allows mobile robot with sensors its gripper to measure generic set features while manipulating an object. We propose switching velocity-force controller grasps object safely and reveals, at the same time, deformation properties. By gently rolling object, can extract additional about contents As application, show use these...

Recognizing useful modular robot configurations composed of hundreds modules is a significant challenge. Matching new configuration to library known essential in identifying and applying control schemes. We present three different algorithms address the problem (a) matching (b) mapping onto configurations. The first method solves using graph isomorphisms can identify that share same underlying structure, but have port connections amongst modules. second approach compares spectra matrices...

Computing a motion that enables mobile manipulator to open door is challenging because it requires tight coordination between the motions of arm and base. Hard-coding motion, on other hand, infeasible since doors vary widely in their sizes types, some are opened by pulling others pushing, indoor spaces often contain obstacles limit freedom degree which up. In this paper, we show how overcome high-dimensionality planning problem identifying graph-based representation small enough for...

Heuristic searches such as the A* search are a popular means of finding least-cost plans due to their generality, strong theoretical guarantees on completeness and optimality, simplicity in implementation consistent behavior. In planning for robotic manipulation, however, these techniques commonly thought impractical high dimensionality problem. this paper, we present heuristic search-based approach motion manipulation that does deal effectively with Our achieves necessary efficiency by...

Much recent work in robot learning from demonstration has focused on automatically segmenting continuous task demonstrations into simpler, reusable primitives.However, strong assumptions are often made about how these primitives can be sequenced, limiting the potential for data reuse.We introduce a novel method discovering semantically grounded and incrementally building improving finite-state representation of which various contingencies arise.Specifically, Beta Process Autoregressive...

Collision-free navigation in cluttered environments is essential for any mobile manipulation system. Traditional systems have relied on a 2D grid map projected from 3D representation efficiency. This approach, however, prevents close to objects situations where configurations are collision within the even if actually no occurs environment. Accordingly, when using such planning paths of robot, number problems which can be solved limited and suboptimal robot may result. We present fast,...

Developing robot agnostic software frameworks involves synthesizing the disparate fields of robotic theory and engineering while simultaneously accounting for a large variability in hardware designs control paradigms. As capabilities increase, setup difficulty learning curve new users also increase. If entry barriers configuring using on robots is too high, even most powerful are useless. A growing need exists to aid getting started with, customizing, framework as necessary...

This paper presents significant steps towards the online integration of 3D perception and manipulation for personal robotics applications. We propose a modular distributed architecture, which seamlessly integrates creation maps collision detection semantic annotations, with real-time motion replanning framework. To validate our system, we present results obtained during comprehensive mobile scenario, includes fusion above components higher level executive.

In this paper, we present a search-based motion planning algorithm for manipulation that handles the high dimensionality of problem and minimizes limitations associated with employing strict set pre-defined actions. Our approach employs adaptive primitives comprised static motions variable on-the-fly generated by two analytical solvers. This method results in slimmer, multi-dimensional lattice offers ability to satisfy goal constraints precision. To validate our approach, used 7DOF...