Currently Part of following projects:
1. Human-Guided Learning and Bench-marking of Robotic Heap Sorting (HEAP) This project will provide scientific advancements for benchmarking, object recognition, manipulation and human-robot interaction.
2. CoRSA: Co-manipulated Robotic Training and Skill Assistance for Telemanipulation in Nuclear Settings ( NCNR flexible partnership funding)
This project aims to develop a novel approach to develop a training facility for teleoperators via providing variable degrees of autonomy while providing robotic assistance. Our approach relies on the adaptive shared control paradigm to teach a human how to better use a teleoperation system.
Earlier Works:
Path Planning of Dual-Arm Space RobotOne of the major areas in space science that demands immediate attention and commercial drive is providing On Orbit Services (OSS) such as refueling of satellites, repair and refurbishment of disabled satellites, active debris removal etc. These autonomous on-orbit services include intricate maneuvers, and hence, use of robots is inevitable for successful completion of these tasks.
|
Reactionless Visual Servoing of Space Robot
An analytical framework for kinematics of a free-floating multi-arm robotic system has been proposed for executing multiple tasks. It is worth mentioning that, while designing image based visual servoing of multi-arm free-floating robotic system, a significant contribution is made in controlling the attitude of the robotic system. A example is provided below.
Initial configuration (in light gray for base and dashed lines for arms) and final configuration (in dark for base and thick lines for arms) after GJM-based visual servoing is shown in (a) and reactionless visual servoing in (b). Plot of norm of pixel error over time for camera-1 and camera-2 is given in (c) for reactionless case. Norm of pixel error for camera-2 is small compared to camera-1, but (c) is still an exponentially decreasing graph. Joint rates of arm-1 and arm-2 is depicted in (d), they decay to zero over time. Plot of base angular position over time is given in (e) and coupling moment between the arms and base is given in (f) for both cases, A and B.
Initial configuration (in light gray for base and dashed lines for arms) and final configuration (in dark for base and thick lines for arms) after GJM-based visual servoing is shown in (a) and reactionless visual servoing in (b). Plot of norm of pixel error over time for camera-1 and camera-2 is given in (c) for reactionless case. Norm of pixel error for camera-2 is small compared to camera-1, but (c) is still an exponentially decreasing graph. Joint rates of arm-1 and arm-2 is depicted in (d), they decay to zero over time. Plot of base angular position over time is given in (e) and coupling moment between the arms and base is given in (f) for both cases, A and B.
Visual Servoing Towards Tumbling ObjectsThe existing approaches aimed to estimate the structure and motion of the tumbling object either explicitly use a object model or employ reconstruction techniques. This motivated the choice of image based visual servoing for capture of tumbling or-
biting objects. It should be noted that unlike common approaches the proposed method does not require estimation of inertia, centroid, angular velocities or orientation of the unknown tumbling object. Video |
Tumbling object real robot experiments on two Pioneer P3-DX robots. Here the tumbling object motion is exhibited by
the first robot having markers on it. (a) Second robot at initial position observes the rotating feature on the first robot. (b) The rotating features are modeled as ellipses, and these ellipses has to move towards left to reach the desired four points. (c) Robot has reached the desired pose. (d) The ellipse features converged to desired stationary points.
the first robot having markers on it. (a) Second robot at initial position observes the rotating feature on the first robot. (b) The rotating features are modeled as ellipses, and these ellipses has to move towards left to reach the desired four points. (c) Robot has reached the desired pose. (d) The ellipse features converged to desired stationary points.
Probabilistic Image Model Based Visual ServoingMajor concern in any IBVS scheme is the uncertainty in environmental conditions substantially generated by unreliable sensor inputs, noise, varying lighting conditions and occlusions. Therefore we analyse the visual servoing schemes through probabilities. We propose a novel approach to visual servoing using mixture models which consider the whole image as a probabilistic function.
|
Random Sampling in Image Space for Path-planning and Execution
Task constraints like image plane limits, joint limits, kinodynamic limits of motors, kinematic, dynamic or algorithmic singularities are other problems inherent to a robotic system while executing a image based visual servoing task. The proposed algorithm uses Rapidly exploring Random Tree (RRT) based sampling in the image feature space for visual servoing of a free-floating multi-arm robotic system to satisfy multiple task constraints.
|
Earth Based Experimental Setup
Development of dual-arm robotic system that replicates zero gravity conditions for planar robots (PDF). Al-though similar systems exist elsewhere, the planar dual-arm space robot we have built is distinctive by being relatively lightweight, compact and modular. The setup can be used to test a wide range of experiments such as visual servoing, reactionless maneuvering and object grasping in space.
|