Pub Date : 2017-11-01DOI: 10.1109/HUMANOIDS.2017.8246882
Shunichi Nozawa, Masaki Murooka, Shintaro Noda, Kunio Kojima, Yuta Kojio, Youhei Kakiuchi, K. Okada, M. Inaba
In the case of object manipulation, a humanoid robot should consider the two-body problem between the object and the robot. To achieve this, the motion planner and the controller must satisfy constraints among the robot, the object, and environments. In addition, the objecfs properties such as mass properties and friction are not known a priori and the robot must obtain this information based on sensor feedback. In this paper, we propose a method for uniform humanoid manipulation of an unknown object by estimating objectenvironment constraints based on changes in the robofs force sensor measurements. The proposed method supports various types of manipulation (lifting, pushing, pivoting), various robot contacts (single-armed, dual-armed, full-body), multi-robot cooperative manipulation, and motion on a movable object. We evaluate the proposed method through experiments involving manipulation of large and heavy objects using life-sized real robots.
{"title":"Unified humanoid manipulation of an object of unknown mass properties and friction based on online constraint estimation","authors":"Shunichi Nozawa, Masaki Murooka, Shintaro Noda, Kunio Kojima, Yuta Kojio, Youhei Kakiuchi, K. Okada, M. Inaba","doi":"10.1109/HUMANOIDS.2017.8246882","DOIUrl":"https://doi.org/10.1109/HUMANOIDS.2017.8246882","url":null,"abstract":"In the case of object manipulation, a humanoid robot should consider the two-body problem between the object and the robot. To achieve this, the motion planner and the controller must satisfy constraints among the robot, the object, and environments. In addition, the objecfs properties such as mass properties and friction are not known a priori and the robot must obtain this information based on sensor feedback. In this paper, we propose a method for uniform humanoid manipulation of an unknown object by estimating objectenvironment constraints based on changes in the robofs force sensor measurements. The proposed method supports various types of manipulation (lifting, pushing, pivoting), various robot contacts (single-armed, dual-armed, full-body), multi-robot cooperative manipulation, and motion on a movable object. We evaluate the proposed method through experiments involving manipulation of large and heavy objects using life-sized real robots.","PeriodicalId":143992,"journal":{"name":"2017 IEEE-RAS 17th International Conference on Humanoid Robotics (Humanoids)","volume":"136 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114213151","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-11-01DOI: 10.1109/HUMANOIDS.2017.8246952
N. Bohorquez, Pierre-Brice Wieber
When a biped robot is walking in a crowd, being able to adapt the duration of steps is a key element to avoid collisions. Model Predictive Control (MPC) schemes for biped walking usually assume a fixed step duration since adapting it leads to a nonlinear problem, in general. Nonlinear solvers do not guarantee the satisfaction of nonlinear constraints at every iterate and this can be problematic for the real-time operation of robots. We propose a method to make sure that all iterates satisfy the nonlinear constraints by borrowing concepts from robust control: we make the problem robust to nonlinearities within some bounds. These bounds are linear with respect to the variables of the problem and can be adapted online.
{"title":"Adaptive step duration in biped walking: A robust approach to nonlinear constraints","authors":"N. Bohorquez, Pierre-Brice Wieber","doi":"10.1109/HUMANOIDS.2017.8246952","DOIUrl":"https://doi.org/10.1109/HUMANOIDS.2017.8246952","url":null,"abstract":"When a biped robot is walking in a crowd, being able to adapt the duration of steps is a key element to avoid collisions. Model Predictive Control (MPC) schemes for biped walking usually assume a fixed step duration since adapting it leads to a nonlinear problem, in general. Nonlinear solvers do not guarantee the satisfaction of nonlinear constraints at every iterate and this can be problematic for the real-time operation of robots. We propose a method to make sure that all iterates satisfy the nonlinear constraints by borrowing concepts from robust control: we make the problem robust to nonlinearities within some bounds. These bounds are linear with respect to the variables of the problem and can be adapted online.","PeriodicalId":143992,"journal":{"name":"2017 IEEE-RAS 17th International Conference on Humanoid Robotics (Humanoids)","volume":"507 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116554486","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-11-01DOI: 10.1109/HUMANOIDS.2017.8246898
Z. Luo, Xuechao Chen, Zhangguo Yu, Qiang Huang, Libo Meng, Qingqing Li, Weimin Zhang, Wenjuan Guo, A. Ming
Increased walking stability and energy efficiency are both important factors for enhancement of the performance of a biped robot. However, it is difficult to derive the optimal control law that is required using optimal control theory because of the strong nonlinearity and the strong coupling of the robot dynamics equation. Use of numerical methods is one effective way to design an optimal control law. This paper presents a method for optimization of the trajectory of a biped robot's swinging leg that is based on a Gaussian pseudospectral method. We first establish a Lagrange optimization function to optimize both the torque and speed during the walking process. By giving different weights to the torque and the speed, optimization of the different targets can be realized, and as a result, a reduction in the velocity can also change the amplitude of the joint motion fluctuations. The effectiveness of the proposed method is demonstrated via simulations and Experiments.
{"title":"Trajectory optimization of humanoid robots swinging leg","authors":"Z. Luo, Xuechao Chen, Zhangguo Yu, Qiang Huang, Libo Meng, Qingqing Li, Weimin Zhang, Wenjuan Guo, A. Ming","doi":"10.1109/HUMANOIDS.2017.8246898","DOIUrl":"https://doi.org/10.1109/HUMANOIDS.2017.8246898","url":null,"abstract":"Increased walking stability and energy efficiency are both important factors for enhancement of the performance of a biped robot. However, it is difficult to derive the optimal control law that is required using optimal control theory because of the strong nonlinearity and the strong coupling of the robot dynamics equation. Use of numerical methods is one effective way to design an optimal control law. This paper presents a method for optimization of the trajectory of a biped robot's swinging leg that is based on a Gaussian pseudospectral method. We first establish a Lagrange optimization function to optimize both the torque and speed during the walking process. By giving different weights to the torque and the speed, optimization of the different targets can be realized, and as a result, a reduction in the velocity can also change the amplitude of the joint motion fluctuations. The effectiveness of the proposed method is demonstrated via simulations and Experiments.","PeriodicalId":143992,"journal":{"name":"2017 IEEE-RAS 17th International Conference on Humanoid Robotics (Humanoids)","volume":"40 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116808263","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-11-01DOI: 10.1109/HUMANOIDS.2017.8246887
Kunihiro Ogata, Hideyuki Tanaka, Y. Matsumoto
Elderly individuals are likely to develop locomotive disorders such as osteoarthritis or osteoporosis. This increases the risk of falls and makes independent movement difficult. Elderly individuals should better understand walking function to extend their healthy life. We therefore propose a new method for estimating hand and foot reaction forces using only visual markers and a monocular camera. When humans contact the environment with their hands, their hand and feet positions define a convex hull. A proposed “ generalized zero moment point ” is projected on this convex hull, which is approximated as a line or plane, and the distance between this point and each contact point is calculated. Reaction forces are calculated based on the ratios of these distances. Evaluation experiments show high agreement between estimated and measured forces of both hands and feet, confirming the validity of the proposed algorithm.
{"title":"Estimating hand and foot reaction forces based on a generalized zero moment point for rehabilitation assist system","authors":"Kunihiro Ogata, Hideyuki Tanaka, Y. Matsumoto","doi":"10.1109/HUMANOIDS.2017.8246887","DOIUrl":"https://doi.org/10.1109/HUMANOIDS.2017.8246887","url":null,"abstract":"Elderly individuals are likely to develop locomotive disorders such as osteoarthritis or osteoporosis. This increases the risk of falls and makes independent movement difficult. Elderly individuals should better understand walking function to extend their healthy life. We therefore propose a new method for estimating hand and foot reaction forces using only visual markers and a monocular camera. When humans contact the environment with their hands, their hand and feet positions define a convex hull. A proposed “ generalized zero moment point ” is projected on this convex hull, which is approximated as a line or plane, and the distance between this point and each contact point is calculated. Reaction forces are calculated based on the ratios of these distances. Evaluation experiments show high agreement between estimated and measured forces of both hands and feet, confirming the validity of the proposed algorithm.","PeriodicalId":143992,"journal":{"name":"2017 IEEE-RAS 17th International Conference on Humanoid Robotics (Humanoids)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124934250","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-11-01DOI: 10.1109/HUMANOIDS.2017.8246910
F. Ledezma, S. Haddadin
Modeling physical systems with neural networks (NN) requires expert architects to determine the best number of nodes, layers and activation functions. For complex systems, such as articulated robots, reported results are limited in accuracy and generalization capabilities. In this work, we introduce the concept FOPnet. It is based on first-order principles and system knowledge to determine topologies of parametrized operator networks that accurately model input-output mappings of physical systems. These topologies consist of meaningful building elements and connections as well as a reduced number of parameters that describe the variables' interdependencies. In this way, learning speed is boosted and the model's accuracy, precision and generalization power improved. We apply the methodology to a 7 degrees-of-freedom LWR4 manipulator and discuss the estimation and generalization capabilities of the network. Results are compared to conventional Feed Forward NN as well as a state-of-the-art Deep Recurrent NN. For the considered complex robot dynamics FOPnet was able to achieve a seven orders of magnitude smaller generalization RMSE.
{"title":"First-order-principles-based constructive network topologies: An application to robot inverse dynamics","authors":"F. Ledezma, S. Haddadin","doi":"10.1109/HUMANOIDS.2017.8246910","DOIUrl":"https://doi.org/10.1109/HUMANOIDS.2017.8246910","url":null,"abstract":"Modeling physical systems with neural networks (NN) requires expert architects to determine the best number of nodes, layers and activation functions. For complex systems, such as articulated robots, reported results are limited in accuracy and generalization capabilities. In this work, we introduce the concept FOPnet. It is based on first-order principles and system knowledge to determine topologies of parametrized operator networks that accurately model input-output mappings of physical systems. These topologies consist of meaningful building elements and connections as well as a reduced number of parameters that describe the variables' interdependencies. In this way, learning speed is boosted and the model's accuracy, precision and generalization power improved. We apply the methodology to a 7 degrees-of-freedom LWR4 manipulator and discuss the estimation and generalization capabilities of the network. Results are compared to conventional Feed Forward NN as well as a state-of-the-art Deep Recurrent NN. For the considered complex robot dynamics FOPnet was able to achieve a seven orders of magnitude smaller generalization RMSE.","PeriodicalId":143992,"journal":{"name":"2017 IEEE-RAS 17th International Conference on Humanoid Robotics (Humanoids)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123762977","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-11-01DOI: 10.1109/HUMANOIDS.2017.8246954
E. Hoffman, A. Rocchi, Arturo Laurenzi, N. Tsagarakis
In this paper we present OpenSoT, an open-source, recently developed software library, that can be used to solve robotics related control problems in a flexible and easy way. OpenSoT includes high-level interfaces to state-of-the-art algorithms for kinematic/dynamic modelling, quadratic programming optimization, cost functions and constraints specification. OpenSoT is implemented in C++ and permits rapid prototyping of controllers for fixed or floating base, highly redundant robots such as (but not limited to) manipulators and humanoids. We discuss the use of OpenSoT from the perspective of the developer and the user, leaving out details on the implementation of the tool. We demonstrate how the software can be used with two examples: control of a redundant humanoid robot through simple inverse kinematics schemes and contact forces optimization.
{"title":"Robot control for dummies: Insights and examples using OpenSoT","authors":"E. Hoffman, A. Rocchi, Arturo Laurenzi, N. Tsagarakis","doi":"10.1109/HUMANOIDS.2017.8246954","DOIUrl":"https://doi.org/10.1109/HUMANOIDS.2017.8246954","url":null,"abstract":"In this paper we present OpenSoT, an open-source, recently developed software library, that can be used to solve robotics related control problems in a flexible and easy way. OpenSoT includes high-level interfaces to state-of-the-art algorithms for kinematic/dynamic modelling, quadratic programming optimization, cost functions and constraints specification. OpenSoT is implemented in C++ and permits rapid prototyping of controllers for fixed or floating base, highly redundant robots such as (but not limited to) manipulators and humanoids. We discuss the use of OpenSoT from the perspective of the developer and the user, leaving out details on the implementation of the tool. We demonstrate how the software can be used with two examples: control of a redundant humanoid robot through simple inverse kinematics schemes and contact forces optimization.","PeriodicalId":143992,"journal":{"name":"2017 IEEE-RAS 17th International Conference on Humanoid Robotics (Humanoids)","volume":"101 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122694924","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-11-01DOI: 10.1109/HUMANOIDS.2017.8246897
U. Scarcia, G. Berselli, G. Palli, C. Melchiorri
In this paper, a novel 3D printed Rotational Joint (RJ) embedding an integrated elastic element is presented. The RJ, produced as a single piece by means of an FDM printer, comprises a traditional pin hinge coupled with a pair of spiral torsion springs, providing the desired compliance for the application at hand. Benefits of the proposed design include monolithic manufacturing and possibility to be successfully employed in robotic articulated devices requiring joint elasticity for their functioning. On the other hand, the sub-optimal RJ behavior, mainly caused by the unavoidable friction between 3D printed mating surfaces, must be accurately taken into account for design purposes. In this context, preliminary reliability tests have been performed showing promising results in terms of lifetime and negligible fatigue effects. Then, a mathematical model of the system is derived, which comprises the spring elasticity along with any frictional effects that may be due to either the pin hinge itself or the tendon transmission (frequently employed in underactuated robotic devices). The model parameters have been empirically evaluated by comparing simulated and experimental data. In addition, the last part of the paper describes how the proposed RJ can be effectively employed for the design of modular, underactuated fingers, providing three degrees of freedom and a single tendon transmission. To this end the model of the joint module proposed in this work will be the starting point for the geometry dimensioning of a finger with a desired free closure motion.
{"title":"Modeling, design, and experimental evaluation of rotational elastic joints for underactuated robotic fingers","authors":"U. Scarcia, G. Berselli, G. Palli, C. Melchiorri","doi":"10.1109/HUMANOIDS.2017.8246897","DOIUrl":"https://doi.org/10.1109/HUMANOIDS.2017.8246897","url":null,"abstract":"In this paper, a novel 3D printed Rotational Joint (RJ) embedding an integrated elastic element is presented. The RJ, produced as a single piece by means of an FDM printer, comprises a traditional pin hinge coupled with a pair of spiral torsion springs, providing the desired compliance for the application at hand. Benefits of the proposed design include monolithic manufacturing and possibility to be successfully employed in robotic articulated devices requiring joint elasticity for their functioning. On the other hand, the sub-optimal RJ behavior, mainly caused by the unavoidable friction between 3D printed mating surfaces, must be accurately taken into account for design purposes. In this context, preliminary reliability tests have been performed showing promising results in terms of lifetime and negligible fatigue effects. Then, a mathematical model of the system is derived, which comprises the spring elasticity along with any frictional effects that may be due to either the pin hinge itself or the tendon transmission (frequently employed in underactuated robotic devices). The model parameters have been empirically evaluated by comparing simulated and experimental data. In addition, the last part of the paper describes how the proposed RJ can be effectively employed for the design of modular, underactuated fingers, providing three degrees of freedom and a single tendon transmission. To this end the model of the joint module proposed in this work will be the starting point for the geometry dimensioning of a finger with a desired free closure motion.","PeriodicalId":143992,"journal":{"name":"2017 IEEE-RAS 17th International Conference on Humanoid Robotics (Humanoids)","volume":"69 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122586840","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-11-01DOI: 10.1109/HUMANOIDS.2017.8239544
T. Petrič, Misel Cevzar, J. Babič
What are the benefits of performing a task with other partners in a physically interactive manipulation task setups? By utilizing a novel human motor learning paradigm, where two individuals are aware of each other and their hands are physically connected through an object, we investigated how each partner adapts his/her motor behavior. We first analyzed performance of twenty subjects on a task where a long object, i.e. a pipe, needs to be manipulated into a groove with different tolerances. We tested efficiency and accuracy of performing the task in two different scenarios: a) one human alone — twenty subjects; b) two humans cooperating — ten pairs. We observed that the task performance during cooperative manipulation of an object does not follow any rules, i.e. either both partners get worse, or both get better, or one partner get and one get worse. By exploiting this properties, we propose a novel control algorithm for robots in physically interactive and cooperative human-robot setups, where the robot adapts to the performance of his/hers partner. This way, it allows the human partner to improve his/hers task performance. The results show that the proposed approach can successfully adapt and match motion of the human partner, and thereby enable the human partner to improve his/her motor skills. After adaption, the human coupled with a robotic partner, can perform the task faster.
{"title":"Utilizing speed-accuracy trade-off models for human-robot coadaptation during cooperative groove fitting task","authors":"T. Petrič, Misel Cevzar, J. Babič","doi":"10.1109/HUMANOIDS.2017.8239544","DOIUrl":"https://doi.org/10.1109/HUMANOIDS.2017.8239544","url":null,"abstract":"What are the benefits of performing a task with other partners in a physically interactive manipulation task setups? By utilizing a novel human motor learning paradigm, where two individuals are aware of each other and their hands are physically connected through an object, we investigated how each partner adapts his/her motor behavior. We first analyzed performance of twenty subjects on a task where a long object, i.e. a pipe, needs to be manipulated into a groove with different tolerances. We tested efficiency and accuracy of performing the task in two different scenarios: a) one human alone — twenty subjects; b) two humans cooperating — ten pairs. We observed that the task performance during cooperative manipulation of an object does not follow any rules, i.e. either both partners get worse, or both get better, or one partner get and one get worse. By exploiting this properties, we propose a novel control algorithm for robots in physically interactive and cooperative human-robot setups, where the robot adapts to the performance of his/hers partner. This way, it allows the human partner to improve his/hers task performance. The results show that the proposed approach can successfully adapt and match motion of the human partner, and thereby enable the human partner to improve his/her motor skills. After adaption, the human coupled with a robotic partner, can perform the task faster.","PeriodicalId":143992,"journal":{"name":"2017 IEEE-RAS 17th International Conference on Humanoid Robotics (Humanoids)","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121281271","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-11-01DOI: 10.1109/HUMANOIDS.2017.8246913
Manuel Baum, Matthew Bernstein, Roberto Martín-Martín, S. Höfer, Johannes Kulick, M. Toussaint, A. Kacelnik, O. Brock
How can we close the gap between animals and robots when it comes to intelligently interacting with the environment? On our quest for answers, we have investigated the problem of physically exploring complex mechanical puzzles, called lockboxes. Biologists have discovered that cockatoos are intrinsically motivated to explore and solve such problems through physical explorative behavior. In this work, we study how different strategies shape the robots' exploration, given basic perception-action skills. Our evaluation highlights the influence of different statistical priors on the performance of the exploration strategies, showing that not only a range of computational methods, but also a range of priors could explain different exploration behaviors. We carry out our study of exploration strategies both in simulation and on two robot platforms. This first step towards a fully integrated real-world system allowed us to identify and remove limitations of our prior theoretical work on cross-entropy-based exploration when applied to complex realistic scenarios. In this paper we propose novel variants of this strategy and our experiments verify that the cross-entropy method performs well on a physical lockbox analogue of the cockatoo apparatus, and can generalize to lockboxes of different properties.
{"title":"Opening a lockbox through physical exploration","authors":"Manuel Baum, Matthew Bernstein, Roberto Martín-Martín, S. Höfer, Johannes Kulick, M. Toussaint, A. Kacelnik, O. Brock","doi":"10.1109/HUMANOIDS.2017.8246913","DOIUrl":"https://doi.org/10.1109/HUMANOIDS.2017.8246913","url":null,"abstract":"How can we close the gap between animals and robots when it comes to intelligently interacting with the environment? On our quest for answers, we have investigated the problem of physically exploring complex mechanical puzzles, called lockboxes. Biologists have discovered that cockatoos are intrinsically motivated to explore and solve such problems through physical explorative behavior. In this work, we study how different strategies shape the robots' exploration, given basic perception-action skills. Our evaluation highlights the influence of different statistical priors on the performance of the exploration strategies, showing that not only a range of computational methods, but also a range of priors could explain different exploration behaviors. We carry out our study of exploration strategies both in simulation and on two robot platforms. This first step towards a fully integrated real-world system allowed us to identify and remove limitations of our prior theoretical work on cross-entropy-based exploration when applied to complex realistic scenarios. In this paper we propose novel variants of this strategy and our experiments verify that the cross-entropy method performs well on a physical lockbox analogue of the cockatoo apparatus, and can generalize to lockboxes of different properties.","PeriodicalId":143992,"journal":{"name":"2017 IEEE-RAS 17th International Conference on Humanoid Robotics (Humanoids)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114928795","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2017-11-01DOI: 10.1109/HUMANOIDS.2017.8246955
Felix Sygulla, Robert Wittmann, Philipp Seiwald, Arne-Christoph Hildebrandt, Daniel Wahrmann, D. Rixen
Traversing uneven terrain with unexpected changes in ground height still poses a major challenge to walking stabilization of humanoid robots. A common approach to balance a biped in such situations is the control of the ground reaction forces at the feet. However, the dynamics of the center of mass is not considered in existing solutions for this direct force control scheme. In this work, we present a force control method to realize contact forces by accelerating the center of mass, which is directly integrated into our hybrid position/force control scheme. For this, we first introduce an analytical formulation for a contact model in task-space. We evaluate the performance of our approach in simulation and real-world experiments with our humanoid robot LOLA. The integration of center of mass dynamics shows great reduction of upper-body inclination angles for a late contact experiment with 5.5 cm change in ground height. We found that by using the system's center of mass dynamics in the force controller, undesired movements along the under-actuated degrees of freedom can be compensated effectively. We consider our approach a starting point for the development of more sophisticated direct force control concepts for humanoid robots.
{"title":"Hybrid position/force control for biped robot stabilization with integrated center of mass dynamics","authors":"Felix Sygulla, Robert Wittmann, Philipp Seiwald, Arne-Christoph Hildebrandt, Daniel Wahrmann, D. Rixen","doi":"10.1109/HUMANOIDS.2017.8246955","DOIUrl":"https://doi.org/10.1109/HUMANOIDS.2017.8246955","url":null,"abstract":"Traversing uneven terrain with unexpected changes in ground height still poses a major challenge to walking stabilization of humanoid robots. A common approach to balance a biped in such situations is the control of the ground reaction forces at the feet. However, the dynamics of the center of mass is not considered in existing solutions for this direct force control scheme. In this work, we present a force control method to realize contact forces by accelerating the center of mass, which is directly integrated into our hybrid position/force control scheme. For this, we first introduce an analytical formulation for a contact model in task-space. We evaluate the performance of our approach in simulation and real-world experiments with our humanoid robot LOLA. The integration of center of mass dynamics shows great reduction of upper-body inclination angles for a late contact experiment with 5.5 cm change in ground height. We found that by using the system's center of mass dynamics in the force controller, undesired movements along the under-actuated degrees of freedom can be compensated effectively. We consider our approach a starting point for the development of more sophisticated direct force control concepts for humanoid robots.","PeriodicalId":143992,"journal":{"name":"2017 IEEE-RAS 17th International Conference on Humanoid Robotics (Humanoids)","volume":"137 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2017-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127425561","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：