In this work, we focus on multi-step manipulation tasks that involve long-horizon planning and considers progress reversal. Such tasks interlace high-level reasoning that consists of the expected states that can be attained to achieve an overall task and low-level reasoning that decides what actions will yield these states. We propose a sample efficient Previous Action Conditioned Robotic Manipulation Network (PAC-RoManNet) to learn the action-value functions and predict manipulation action candidates from visual observation of the scene and action-value predictions of the previous action. We define a Task Progress based Gaussian (TPG) reward function that computes the reward based on actions that lead to successful motion primitives and progress towards the overall task goal. To balance the ratio of exploration/exploitation, we introduce a Loss Adjusted Exploration (LAE) policy that determines actions from the action candidates according to the Boltzmann distribution of loss estimates. We demonstrate the effectiveness of our approach by training PAC-RoManNet to learn several challenging multi-step robotic manipulation tasks in both simulation and real-world. Experimental results show that our method outperforms the existing methods and achieves state-of-the-art performance in terms of success rate and action efficiency. The ablation studies show that TPG and LAE are especially beneficial for tasks like multiple block stacking. Additional experiments on Ravens-10 benchmark tasks suggest good generalizability of the proposed PAC-RoManNet.
Multi-step manipulation tasks in unstructured environments are extremely challenging for a robot to learn. Such tasks interlace high-level reasoning that consists of the expected states that can be attained to achieve an overall task and low-level reasoning that decides what actions will yield these states. We propose a model-free deep reinforcement learning method to learn these multi-step manipulation tasks. We introduce a Robotic Manipulation Network (RoManNet) which is a vision-based deep reinforcement learning algorithm to learn the action-value functions and project manipulation action candidates. We define a Task Progress based Gaussian (TPG) reward function that computes the reward based on actions that lead to successful motion primitives and progress towards the overall task goal. We further introduce a Loss Adjusted Exploration (LAE) policy that determines actions from the action candidates according to the Boltzmann distribution of loss estimates. We demonstrate the effectiveness of our approaches by training RoManNet to learn several challenging multi-step robotic manipulation tasks. Empirical results show that our method outperforms the existing methods and achieves state-of-the-art results. The ablation studies show that TPG and LAE are especially beneficial for tasks like multiple block stacking. Code is available at: this https URL
This paper presents the design and development of a low cost and user friendly interface for the control of a 6-DOF slave tele-operated anthropomorphic robotic arm. Articulation of the robotic arm is achieved about six single-axis revolute joints: one for each shoulder abduction-adduction (abd-add), shoulder flexion-extension (flx-ext), elbow flx-ext, wrist flx-ext, wrist radial-ulnar (rad-uln), and gripper open-close. Tele-operator, master, uses the Man Machine Interface (MMI) to operate in real-time the robotic arm. The MMI has simple motion capture devices that translate motion into analog voltages which bring about the corresponding actuating signals in the robotic arm.
In this work, we present a deep reinforcement learning based method to solve the problem of robotic grasping using visio-motor feedback. The use of a deep learning based approach reduces the complexity caused by the use of hand-designed features. Our method uses an off-policy reinforcement learning framework to learn the grasping policy. We use the double deep Q-learning framework along with a novel GraspQ-Network to output grasp probabilities used to learn grasps that maximize the pick success. We propose a visual servoing mechanism that uses a multi-view camera setup that observes the scene which contains the objects of interest. We performed experiments using a Baxter Gazebo simulated environment as well as on the actual robot. The results show that our proposed method outperforms the baseline Q-learning framework and increases grasping accuracy by adapting a multi-view model in comparison to a single-view model.
In robotics, there is need of an interactive and expedite learning method as experience is expensive. Robot Learning from Demonstration (RLfD) enables a robot to learn a policy from demonstrations performed by teacher. RLfD enables a human user to add new capabilities to a robot in an intuitive manner, without explicitly reprogramming it. In this work, we present a novel interactive framework, where a collaborative robot learns skills for trajectory based tasks from demonstrations performed by a human teacher. The robot extracts features from each demonstration called as key-points and learns a model of the demonstrated skill using Hidden Markov Model (HMM). Our experimental results show that the learned model can be used to produce a generalized trajectory based skill.
Through this paper, we elucidate the second phase of the design and development of the tele-operated humanoid robot Dexto:Eka:. Phase one comprised of the development of a 6 DoF left anthropomorphic arm and left exo-frame. Here, we illustrate the development of the right arm, right exo-frame, torso, backbone, human machine interface and omni-directional locomotion system. Dexto:Eka: will be able to communicate with a remote user through Wi-Fi. An exo-frame capacitates it to emulate human arms and its locomotion is controlled by joystick. A Graphical User Interface monitors and helps in controlling the system.
In this paper, we present a modular robotic system to tackle the problem of generating and performing antipodal robotic grasps for unknown objects from n-channel image of the scene. We propose a novel Generative Residual Convolutional Neural Network (GR-ConvNet) model that can generate robust antipodal grasps from n-channel input at real-time speeds (~20ms). We evaluate the proposed model architecture on standard datasets and a diverse set of household objects. We achieved state-of-the-art accuracy of 97.7% and 94.6% on Cornell and Jacquard grasping datasets respectively. We also demonstrate a grasp success rate of 95.4% and 93% on household and adversarial objects respectively using a 7 DoF robotic arm.
The presented paper is concerned with designing of a low-cost, easy to use, intuitive interface for the control of a slave anthropomorphic teleo- operated robot. Tele-operator “masters”, that operate in real-time with the robot, have ranged from simple motion capture devices, to more complex force reflective exoskeletal masters. Our general design approach has been to begin with the definition of desired objective behaviours, rather than the use of available components with their predefined technical specifications. With the technical specifications of the components necessary to achieve the desired behaviours defined, the components are either acquired, or in most cases, developed and built. The control system, which includes the operation of feedback approaches, acting in collaboration with physical machinery, is then defined and implemented.
In this work, we present a deep reinforcement learning based method to solve the problem of robotic grasping using visio-motor feedback. The use of a deep learning based approach reduces the complexity caused by the use of hand-designed features. Our method uses an off-policy reinforcement learning framework to learn the grasping policy. We use the double deep Q-learning framework along with a novel Grasp-Q-Network to output grasp probabilities used to learn grasps that maximize the pick success. We propose a visual servoing mechanism that uses a multi-view camera setup that observes the scene which contains the objects of interest. We performed experiments using a Baxter Gazebo simulated environment as well as on the actual robot. The results show that our proposed method outperforms the baseline Q-learning framework and increases grasping accuracy by adapting a multi-view model in comparison to a single-view model.
In this paper, we presented the Balance model as a singular axis self balancing robot that is capable of adjusting itself with respect to changes in weight and position. We developed the Balance System from a single servo and a single accelerometer. The stability of the system is to show the capabilities of the ATMega8 in doing PID loops even with limited accuracy in position readings. PID control system is designed to monitor the motors so as to keep the system in equilibrium. It should be easily reproducible given the right parts and code.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)
