Search

To link to full-text access for this article, visit this link: http://dx.doi.org/10.1016/j.robot.2007.11.002 Byline: Borhan Beigzadeh (a)(b), Majid Nili Ahmadabadi (b), Ali Meghdari (a), Adel Akbarimajd (b) Abstract: In this paper, we aim at an integrated approach to Dynamic Biped Walking (DBW) and Dynamic Object Manipulation (DOM) at an abstract level. To this end, we offer a unified and abstract concept with a dual interpretation as a DOM and as a DBW system. We validate the proposed approach by using a set of simulations on an illustrative case study and show how it can be used in modeling as well as design of planning and control algorithms for DOM and DBW systems. In the case study, we describe the proposed approach and show its dual interpretation by identifying the relations between 2D dynamic object manipulation of a disc using two planar manipulators and 2D dynamic object locomotion of lower part of a biped robot. More specifically, having obtained the equations of DOM, we change the boundary conditions of the problem in such a way that both radius and mass of the disc tend to infinity. Simultaneously, both size and mass of the manipulators' base, i.e. the planet earth, tend to some values in the order of human body mass and dimension. Regarding these changes, we can transform DOM into DBW and vice versa. To test the proposed approach, a simple control strategy is introduced to handle impact between the manipulators (legs) and the object (the earth). In addition, a motion planning system is designed in such a way that the manipulators (legs) catch and throw the manipulated object (the earth) in appropriate configurations. Author Affiliation: (a) Centre of Excellence in Design, Robotics, and Automation (CEDRA), ME Department, Sharif University of Technology, Tehran, Iran (b) Control and Intelligent Processing Centre of Excellence, Robotics and AI Laboratory, ECE Department, University of Tehran, Tehran, Iran Article History: Received 12 July 2007; Revised 30 October 2007; Accepted 13 November 2007

To link to full-text access for this article, visit this link: http://dx.doi.org/10.1016/j.robot.2007.01.007 Byline: Adel Akbarimajd (a), Majid Nili Ahmadabadi (a), Borhan Beigzadeh (b) Abstract: Dynamic manipulation of polygonal objects by an array of one degree of freedom arms is studied from kinematics and planning points of view. In the studied manipulation method, an object is manipulated to its goal configuration by a sequence of juggles. A kinematic model of an object throwing task is driven and a method for object manipulation by a sequence of throws is proposed. The method, called Backward Throws Method (BTM), is based on throwing an object backward towards the arm pivot. Based on the developed model, a planning algorithm is proposed for BTM. In addition, the only existing similar method to BTM, which is based on forward throws-named FTM in this paper-is reformulated for implementation by a series of arms and compared with the proposed method. Analytical investigations, simulation results, and experimental outcomes show that BTM meets the desired requirements. Moreover, in comparison to FTM, BTM requires fewer number of throws and lower release velocity. In addition, the object's maximum height of flight is much lower in BTM which results in lower catching impact. According to the experimental results, although the proposed method has no feedback from object position, accumulated position error is very small. This fact is directly related to the attained decrease in catching impact which causes small object slippage and rebound on catching. Furthermore, there is no restriction on the arm geometry in BTM while in FTM an arm with negative offset is needed. Author Affiliation: (a) Control and Intelligent Processing Center of Excellence, Robotics and AI Lab., ECE Department, University of Tehran, Iran (b) Center of Excellence in Design, Robotics and Automation (CEDRA), ME Department, Sharif University of Technology, Iran Article History: Received 30 May 2006; Revised 11 January 2007; Accepted 25 January 2007

• Multi-Loss Snapshot Ensemble (MLSE) of CNNs offers feature ensemble learning for OSV. • MLSE proposes simultaneous use of multi-loss function within a sequential training. • MLSE combines advantage of diversity and regularization to tackle challenges of OSV. • USMG-SVM combines decisions by selecting most generalizable SVM for each user. • MLSE + USMG-SVM achieved significant improvements over state-of-the-arts in OSV. Offline Signature Verification (OSV) is a challenging pattern recognition task, especially in the presence of skilled forgeries that are not available during training. This study aims to tackle its challenges and meet the substantial need for generalization for OSV by examining different loss functions for Convolutional Neural Network (CNN). We adopt our new approach to OSV by asking two questions: 1. which classification loss provides more generalization for feature learning in OSV?, and 2. How integration of different losses into a unified multi-loss function lead to an improved learning framework? These questions are studied based on analysis of three loss functions, including cross entropy, Cauchy-Schwarz divergence, and hinge loss. According to complementary features of these losses, we combine them into a dynamic multi-loss function and propose a novel ensemble framework for simultaneous use of them in CNN. Our proposed Multi-Loss Snapshot Ensemble (MLSE) consists of several sequential trials. In each trial, a dominant loss function is selected from the multi-loss set, and the remaining losses act as a regularizer. Different trials learn diverse representations for each input based on signature identification task. This multi-representation set is then employed for the verification task. An ensemble of SVMs is trained on these representations, and their decisions are finally combined according to the selection of most generalizable SVM for each user. We conducted two sets of experiments based on two different protocols of OSV, i.e., writer-dependent and writer-independent on three signature datasets: GPDS-Synthetic, MCYT, and UT-SIG. Based on the writer-dependent OSV protocol, On UT-SIG, we achieved 6.17% Equal Error Rate (EER) which showed substantial improvement over the best EER in the literature, 9.61%. Our method surpassed state-of-the-arts by 2.5% on GPDS-Synthetic, achieving 6.13%. Our result on MCYT was also comparable to the best previous results. The second set of experiments examined the robustness of our proposed method to the arrival of new users enrolled in the OSV system based on the writer-independent protocol. The results also confirmed that our proposed system efficiently performed the verification of new users enrolled in the OSV system. Image, graphical abstract [ABSTRACT FROM AUTHOR]

Abstract There are some biological evidences showing that the actuation system in legged animals is impulsive; it is not continuous. As opposed to continuous control/actuation, the control actions occur in specific intervals, and from the instant of one actuation until the start of the next one, passive elements guarantee the stability of the robotic system and govern its natural dynamics. In this paper, we present an analytical method for concurrent design of impulsive controller and passive elements (compliance and damper) for robotic systems; e.g., manipulators and legged-robots. To optimize the force profiles of passive elements, three different cost functions are presented which optimize the natural dynamics and energy consumption of the robot. The presented method can be applied to both cyclic and non-cyclic (explosive) tasks so as to attain energy efficient and bio-inspired motions. The method is applied to three biological models: a simulated human arm for throwing an object, a swing leg for drawing an oval, and a 3D quadruped robot for performing walking gait. Our findings in the simulation studies are in line with the hypothesis of impulsive actuation in nature and show the applicability of our method in robotics. [ABSTRACT FROM AUTHOR]

Inverse reinforcement learning ( IRL ) is a powerful tool for teaching by demonstrations, provided that sufficiently diverse and optimal demonstrations are given, and learner agent correctly perceives those demonstrations. These conditions are hard to meet in practice; as a trainer cannot cover all possibilities by demonstrations, he may partially fail to follow the optimal behavior. Also, trainer and learner have different perceptions of the environment including trainer's actions. A practical way to overcome these problems is using a combination of trainer's demonstrations and feedbacks. We propose an interactive learning approach to overcome the challenge of non-optimal demonstrations by integrating human evaluative feedbacks with the IRL process, given sufficiently diverse demonstrations and the domain transition model. To this end, we develop a probabilistic model of human feedbacks and iteratively improve the agent policy using Bayes rule. We then integrate this information in an extended IRL algorithm to enhance the learned reward function. We examine the developed approach in one experimental and two simulated tasks; i.e., a grid world navigation, a highway car driving system and a navigation task by the e-puck robot. Obtained results show significant improved efficiency of the proposed approach in face of having different levels of non-optimality in demonstrations and the number of evaluative feedbacks. [ABSTRACT FROM AUTHOR]

We present a novel adaptive oscillator, called Adaptive Natural Oscillator (ANO), to exploit the natural dynamics of a given robotic system. This tool is built upon the Adaptive Frequency Oscillator (AFO), and it can be used as a pattern generator in robotic applications such as locomotion systems. In contrast to AFO, that adapts to the frequency of an external signal, ANO adapts the frequency of reference trajectory to the natural dynamics of the given system. In this work, we prove that, in linear systems, ANO converges to the system’s natural frequency. Furthermore, we show that this tool exploits the natural dynamics for energy efficiency through minimization of actuator effort. This property makes ANO an appealing tool for energy consumption reduction in cyclic tasks; especially in legged systems. We also extend the proposed adaptation mechanism to high dimensional and general cases; such as n -DOF manipulators. In addition, by investigating a hopper leg in simulation, we show the efficacy of ANO in face of dynamical discontinuities; such as those inherent in legged locomotion. Furthermore, we apply ANO to a simulated compliant robotic manipulator performing a periodic task where the energy consumption is drastically reduced. Finally, the experimental results on a 1 -DOF compliant joint show that our adaptive oscillator, despite all practical uncertainties and deviations from theoretical models, exploits the natural dynamics and reduces the energy consumption. [ABSTRACT FROM AUTHOR]

This paper is concerned with the presentation of a parallel compliance adaptation method for systems equipped with rotary motion mechanisms towards obtaining energy efficiency in cyclic tasks over a reasonable range of task frequency variations. In this work, we first introduce a variable parallel elastic actuator (VPEA) design for implementation on uni-directional joints that can respond in line with the torque requirements caused by frequency variations in rotary mechanisms. Then, in the next step, we propose two design approaches namely " general method " and " frequency-based method " for the VPEA along with the stiffness adjustment approaches both in offline and online manners. The optimality and convergence of the adaptation method for the proposed rotary VPEA are also analytically proved in general to be globally exponentially stable in the sense of Lyapunov. Finally, to demonstrate the applicability and efficiency of our VPEA, we deployed it in a robotic leg model as the case study. The simulation results demonstrate the stability and convergence of our adaptation rule and highlight the performance of the proposed VPEA in increasing energy efficiency over a wide range of task frequency variations. [ABSTRACT FROM AUTHOR]