In this paper we introduce the Generalized Contact Control Framework (GCCF) implemented on a compliant robotic manipulator. We demonstrate that the combined joint compliance and GCCF-based compliance control enable the completion of complex contact tasks in uncertain environments, where complex refers to the need to meet different contact force requirements involving multiple steps and output axes. Operating in uncertain environments means limited knowledge of the location or material properties of contact objects. The demonstrated tasks include opening a pill bottle and rigidly connecting to a purely mechanical tool changer. The GCCF simplifies the definition and modification of contact control parameters and allows for on-the-fly definition and completion of new tasks. Unlike hybrid force/impedance controllers, we do not need to define large damping and stiffness matrices, and we decouple the joint level control gains from the compliance control. The result is a robotic manipulator that can dynamically switch between unconstrained motion and contact tasks and provides a lot of versatility to perform a wide variety of tasks.
{"title":"Completing Complex Contact Tasks Using Integrated Active and Passive Compliant Control Methodologies","authors":"Adam Pettinger, M. Pryor","doi":"10.1115/dscc2019-9062","DOIUrl":"https://doi.org/10.1115/dscc2019-9062","url":null,"abstract":"\u0000 In this paper we introduce the Generalized Contact Control Framework (GCCF) implemented on a compliant robotic manipulator. We demonstrate that the combined joint compliance and GCCF-based compliance control enable the completion of complex contact tasks in uncertain environments, where complex refers to the need to meet different contact force requirements involving multiple steps and output axes. Operating in uncertain environments means limited knowledge of the location or material properties of contact objects. The demonstrated tasks include opening a pill bottle and rigidly connecting to a purely mechanical tool changer. The GCCF simplifies the definition and modification of contact control parameters and allows for on-the-fly definition and completion of new tasks. Unlike hybrid force/impedance controllers, we do not need to define large damping and stiffness matrices, and we decouple the joint level control gains from the compliance control. The result is a robotic manipulator that can dynamically switch between unconstrained motion and contact tasks and provides a lot of versatility to perform a wide variety of tasks.","PeriodicalId":41412,"journal":{"name":"Mechatronic Systems and Control","volume":null,"pages":null},"PeriodicalIF":0.6,"publicationDate":"2019-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82622564","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}
We introduce the Restricted Newton’s Method (RNM), a basic optimization method, to accelerate model predictive control turnaround times. RNM is a hybrid of Newton’s method (NM) and gradient descent (GD) that can be used as a building block in nonlinear programming. The two parameters of RNM are the subspace on which we restrict the Newton steps and the maximal size of the GD step. We present a convergence analysis of RNM and demonstrate how these parameters can be selected for MPC applications using simple machine learning methods. This leads to two parameter selection strategies with different convergence behaviour. Lastly, we demonstrate the utility of RNM on a sample autonomous vehicle problem with promising results.
{"title":"The Restricted Newton Method for Fast Nonlinear Model Predictive Control","authors":"A. Maitland, C. Jin, J. McPhee","doi":"10.1115/dscc2019-9067","DOIUrl":"https://doi.org/10.1115/dscc2019-9067","url":null,"abstract":"\u0000 We introduce the Restricted Newton’s Method (RNM), a basic optimization method, to accelerate model predictive control turnaround times. RNM is a hybrid of Newton’s method (NM) and gradient descent (GD) that can be used as a building block in nonlinear programming. The two parameters of RNM are the subspace on which we restrict the Newton steps and the maximal size of the GD step. We present a convergence analysis of RNM and demonstrate how these parameters can be selected for MPC applications using simple machine learning methods. This leads to two parameter selection strategies with different convergence behaviour. Lastly, we demonstrate the utility of RNM on a sample autonomous vehicle problem with promising results.","PeriodicalId":41412,"journal":{"name":"Mechatronic Systems and Control","volume":null,"pages":null},"PeriodicalIF":0.6,"publicationDate":"2019-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81049204","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}
S. T. Payne, C. Garrison, Steve Markham, Tucker Hermans, K. Leang
This paper focuses on the assembly planning process for constructing polygonal furniture (such as cabinets, speakers, bookshelves, etc.) using robotic arms and manipulators. An algorithm is described that utilizes easily-implemented and generally-accepted motion planning algorithms to take advantage of the polygonal nature of the furniture, which reduces the complexity of the assembly planner. In particular, the algorithm disassembles a given CAD model in simulation to find a valid assembly order and disassembly path, then implements that assembly order with two robotic arms, using the disassembly path as the finishing path of the part into the assembly. Additionally, it finds a collision-free plan developed for each of the arms in the correct assembly order with the final result being the assembly of the model.
{"title":"Assembly Planning Using a Two-Arm System for Polygonal Furniture","authors":"S. T. Payne, C. Garrison, Steve Markham, Tucker Hermans, K. Leang","doi":"10.1115/dscc2019-9173","DOIUrl":"https://doi.org/10.1115/dscc2019-9173","url":null,"abstract":"\u0000 This paper focuses on the assembly planning process for constructing polygonal furniture (such as cabinets, speakers, bookshelves, etc.) using robotic arms and manipulators. An algorithm is described that utilizes easily-implemented and generally-accepted motion planning algorithms to take advantage of the polygonal nature of the furniture, which reduces the complexity of the assembly planner. In particular, the algorithm disassembles a given CAD model in simulation to find a valid assembly order and disassembly path, then implements that assembly order with two robotic arms, using the disassembly path as the finishing path of the part into the assembly. Additionally, it finds a collision-free plan developed for each of the arms in the correct assembly order with the final result being the assembly of the model.","PeriodicalId":41412,"journal":{"name":"Mechatronic Systems and Control","volume":null,"pages":null},"PeriodicalIF":0.6,"publicationDate":"2019-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90812082","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}
J. S. Terry, J. Whitaker, R. Beard, Marc D. Killpack
The compliance and other nonlinear dynamics of large-scale soft robots makes effective control difficult. This is especially true when working with unknown payloads or when the system dynamics change over time which is likely to happen for soft robots. In this paper, we present a novel method of coupling model reference adaptive control (MRAC) with model predictive control (MPC) for platforms with antagonistic pneumatic actuators. We demonstrate its utility on a fully inflatable, six degree-of-freedom pneumatically actuated soft robot manipulator that is over two meters long. Specifically, we compare control performance with no integral controller, with an integral controller, and with MRAC when running a nominal model predictive controller with significant weight attached to the end effector.
{"title":"Adaptive Control of Large-Scale Soft Robot Manipulators With Unknown Payloads","authors":"J. S. Terry, J. Whitaker, R. Beard, Marc D. Killpack","doi":"10.1115/dscc2019-9037","DOIUrl":"https://doi.org/10.1115/dscc2019-9037","url":null,"abstract":"\u0000 The compliance and other nonlinear dynamics of large-scale soft robots makes effective control difficult. This is especially true when working with unknown payloads or when the system dynamics change over time which is likely to happen for soft robots. In this paper, we present a novel method of coupling model reference adaptive control (MRAC) with model predictive control (MPC) for platforms with antagonistic pneumatic actuators. We demonstrate its utility on a fully inflatable, six degree-of-freedom pneumatically actuated soft robot manipulator that is over two meters long. Specifically, we compare control performance with no integral controller, with an integral controller, and with MRAC when running a nominal model predictive controller with significant weight attached to the end effector.","PeriodicalId":41412,"journal":{"name":"Mechatronic Systems and Control","volume":null,"pages":null},"PeriodicalIF":0.6,"publicationDate":"2019-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89208145","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}
Multi Actuator Technology was unveiled in December 2017 by Seagate, a breakthrough that can double the data performance of the future generation hard disk drives. This technology will equip drives with dual actuators operating on the same pivot point. Each actuator will control half of the drives arms. This new technology brings new control challenges with it. Since two actuators operate independently on the same pivot timber, the control forces and torques generated by one actuator can affect the operation of the other actuator. The independent functioning of the two actuators will lead to a scenario of one actuator in the track seeking mode while the other actuator is in the track following mode. It is expected that the track seeking actuator will impart vibration onto the track following actuator, worsening its performance drastically. In this paper, we propose a methodology to estimate this imparted vibration and to design feedforward controllers for the voice coil motor and the micro actuator, of the track following actuator, to suppress the estimated vibration. The vibration estimation is performed using power spectral factorization techniques. Whereas, the feedforward controllers are designed using a mixed H2 – H∞ data driven methodology to obtain a robust design.
{"title":"Active Vibration Rejection in Multi Actuator Drives: Data Driven Approach","authors":"Prateek Shah, R. Horowitz","doi":"10.1115/dscc2019-8983","DOIUrl":"https://doi.org/10.1115/dscc2019-8983","url":null,"abstract":"\u0000 Multi Actuator Technology was unveiled in December 2017 by Seagate, a breakthrough that can double the data performance of the future generation hard disk drives. This technology will equip drives with dual actuators operating on the same pivot point. Each actuator will control half of the drives arms.\u0000 This new technology brings new control challenges with it. Since two actuators operate independently on the same pivot timber, the control forces and torques generated by one actuator can affect the operation of the other actuator. The independent functioning of the two actuators will lead to a scenario of one actuator in the track seeking mode while the other actuator is in the track following mode. It is expected that the track seeking actuator will impart vibration onto the track following actuator, worsening its performance drastically.\u0000 In this paper, we propose a methodology to estimate this imparted vibration and to design feedforward controllers for the voice coil motor and the micro actuator, of the track following actuator, to suppress the estimated vibration. The vibration estimation is performed using power spectral factorization techniques. Whereas, the feedforward controllers are designed using a mixed H2 – H∞ data driven methodology to obtain a robust design.","PeriodicalId":41412,"journal":{"name":"Mechatronic Systems and Control","volume":null,"pages":null},"PeriodicalIF":0.6,"publicationDate":"2019-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86659132","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}
In this paper, the performance of a bat searching algorithm is studied from system dynamics point of view. Bat searching algorithm (BA) is a recently developed swarm intelligence based optimization algorithm which has shown great success when solving complicated optimization problems. Each bat in the BA has two main states: velocity and position. The position represents the solution of the optimization problems while the velocity represents the searching direction and step size during each iteration. Due to the nature of the update equations, the dynamics of the bats are formulated as a group of second-order discrete-time systems. In this paper, the performance of the algorithm is analyzed based on the nature of the responses in the second-order systems. The over-damped response, under-damped responses are studied and the parameters requirements are derived. Moreover, unstable scenarios of the bats are also considered when examining the performance of the algorithm. Numerical evaluations are conducted to test different choices of the parameters in the BA.
{"title":"Performance Study of a Bat Searching Algorithm From System Dynamics Perspective","authors":"Haopeng Zhang, N. Schutte","doi":"10.1115/dscc2019-9017","DOIUrl":"https://doi.org/10.1115/dscc2019-9017","url":null,"abstract":"\u0000 In this paper, the performance of a bat searching algorithm is studied from system dynamics point of view. Bat searching algorithm (BA) is a recently developed swarm intelligence based optimization algorithm which has shown great success when solving complicated optimization problems. Each bat in the BA has two main states: velocity and position. The position represents the solution of the optimization problems while the velocity represents the searching direction and step size during each iteration. Due to the nature of the update equations, the dynamics of the bats are formulated as a group of second-order discrete-time systems. In this paper, the performance of the algorithm is analyzed based on the nature of the responses in the second-order systems. The over-damped response, under-damped responses are studied and the parameters requirements are derived. Moreover, unstable scenarios of the bats are also considered when examining the performance of the algorithm. Numerical evaluations are conducted to test different choices of the parameters in the BA.","PeriodicalId":41412,"journal":{"name":"Mechatronic Systems and Control","volume":null,"pages":null},"PeriodicalIF":0.6,"publicationDate":"2019-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88686602","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}
We formulate a predictor-based controller for a high-DOF manipulator to compensate a time-invariant input delay during a pick-and-place task. Robot manipulators are widely used in tele-manipulation systems on the account of their reliable, fast, and precise motions while they are subject to large delays. Using common control algorithms on such delay systems can cause not only poor control performance, but also catastrophic instability in engineering applications. Therefore, delays need to be compensated in designing robust control laws. As a case study, we focus on a 7-DOF Baxter manipulator subject to three different input delays. First, delay-free dynamic equations of the Baxter manipulator are derived using the Lagrangian method. Then, we formulate a predictor-based controller, in the presence of input delay, in order to track desired trajectories. Finally, the effects of input delays in the absence of a robust predictor are investigated, and then the performance of the predictor-based controller is experimentally evaluated to reveal robustness of the algorithm formulated. Simulation and experimental results demonstrate that the predictor-based controller effectively compensates input delays and achieves closed-loop stability.
{"title":"Time Delay Control of a High-DOF Robot Manipulator Through Feedback Linearization Based Predictor","authors":"M. Bagheri, P. Naseradinmousavi, M. Krstić","doi":"10.1115/dscc2019-8915","DOIUrl":"https://doi.org/10.1115/dscc2019-8915","url":null,"abstract":"\u0000 We formulate a predictor-based controller for a high-DOF manipulator to compensate a time-invariant input delay during a pick-and-place task. Robot manipulators are widely used in tele-manipulation systems on the account of their reliable, fast, and precise motions while they are subject to large delays. Using common control algorithms on such delay systems can cause not only poor control performance, but also catastrophic instability in engineering applications. Therefore, delays need to be compensated in designing robust control laws. As a case study, we focus on a 7-DOF Baxter manipulator subject to three different input delays. First, delay-free dynamic equations of the Baxter manipulator are derived using the Lagrangian method. Then, we formulate a predictor-based controller, in the presence of input delay, in order to track desired trajectories. Finally, the effects of input delays in the absence of a robust predictor are investigated, and then the performance of the predictor-based controller is experimentally evaluated to reveal robustness of the algorithm formulated. Simulation and experimental results demonstrate that the predictor-based controller effectively compensates input delays and achieves closed-loop stability.","PeriodicalId":41412,"journal":{"name":"Mechatronic Systems and Control","volume":null,"pages":null},"PeriodicalIF":0.6,"publicationDate":"2019-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89030140","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}
Alex Bertino, M. Bagheri, M. Krstić, P. Naseradinmousavi
In this paper, we examine the autonomous operation of a high-DOF robot manipulator. We investigate a pick-and-place task where the position and orientation of an object, an obstacle, and a target pad are initially unknown and need to be autonomously determined. In order to complete this task, we employ a combination of computer vision, deep learning, and control techniques. First, we locate the center of each item in two captured images utilizing HSV-based scanning. Second, we utilize stereo vision techniques to determine the 3D position of each item. Third, we implement a Convolutional Neural Network in order to determine the orientation of the object. Finally, we use the calculated 3D positions of each item to establish an obstacle avoidance trajectory lifting the object over the obstacle and onto the target pad. Through the results of our research, we demonstrate that our combination of techniques has minimal error, is capable of running in real-time, and is able to reliably perform the task. Thus, we demonstrate that through the combination of specialized autonomous techniques, generalization to a complex autonomous task is possible.
{"title":"Experimental Autonomous Deep Learning-Based 3D Path Planning for a 7-DOF Robot Manipulator","authors":"Alex Bertino, M. Bagheri, M. Krstić, P. Naseradinmousavi","doi":"10.1115/dscc2019-8951","DOIUrl":"https://doi.org/10.1115/dscc2019-8951","url":null,"abstract":"\u0000 In this paper, we examine the autonomous operation of a high-DOF robot manipulator. We investigate a pick-and-place task where the position and orientation of an object, an obstacle, and a target pad are initially unknown and need to be autonomously determined. In order to complete this task, we employ a combination of computer vision, deep learning, and control techniques. First, we locate the center of each item in two captured images utilizing HSV-based scanning. Second, we utilize stereo vision techniques to determine the 3D position of each item. Third, we implement a Convolutional Neural Network in order to determine the orientation of the object. Finally, we use the calculated 3D positions of each item to establish an obstacle avoidance trajectory lifting the object over the obstacle and onto the target pad. Through the results of our research, we demonstrate that our combination of techniques has minimal error, is capable of running in real-time, and is able to reliably perform the task. Thus, we demonstrate that through the combination of specialized autonomous techniques, generalization to a complex autonomous task is possible.","PeriodicalId":41412,"journal":{"name":"Mechatronic Systems and Control","volume":null,"pages":null},"PeriodicalIF":0.6,"publicationDate":"2019-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77559539","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}
This work deals with amplitude frequency response of MEMS cantilever resonators undergoing superharmonic resonance of third order. The cantilever resonator is parallel to a ground plate and under alternating current (AC) voltage that excites the cantilever into vibrations. The driving frequency of the AC voltage is near one sixth of the first natural frequency of the cantilever beam resulting into superharmonic resonance of third order. The cantilever beam is modeled using Euler-Bernoulli beam theory. The electrostatic force is modeled using Palmer’s formula to include the fringe effect. In order to investigate the amplitude frequency behavior of the system reduced order models (ROMs) are developed. Three methods are used to solve these ROMs they are 1) the method of multiple scales (MMS) for ROM with one mode of vibration, 2) homotopy analysis method (HAM) for ROM with one mode of vibration, and 3) direct numerical integration for 2 modes of vibration Reduced Order Model (2T ROM) producing time responses of the tip of the cantilever resonator. In this work the limitations of MMS and HAM are highlighted when considering large voltage values i.e hard excitations. For large voltage values MMS and HAM cannot accurately predict the amplitude frequency response; the results from 2T ROM time responses disagree significantly with the MMS and HAM solutions. The effect of voltage on the frequency response is investigated. As the voltage values in the system increase the responses shift to lower frequencies and larger amplitudes.
{"title":"Amplitude-Frequency Response of Superharmonic Resonance of Third Order of Electrostatically Actuated MEMS Cantilever Resonators","authors":"D. Caruntu, Christian Reyes","doi":"10.1115/dscc2019-9172","DOIUrl":"https://doi.org/10.1115/dscc2019-9172","url":null,"abstract":"\u0000 This work deals with amplitude frequency response of MEMS cantilever resonators undergoing superharmonic resonance of third order. The cantilever resonator is parallel to a ground plate and under alternating current (AC) voltage that excites the cantilever into vibrations. The driving frequency of the AC voltage is near one sixth of the first natural frequency of the cantilever beam resulting into superharmonic resonance of third order. The cantilever beam is modeled using Euler-Bernoulli beam theory. The electrostatic force is modeled using Palmer’s formula to include the fringe effect. In order to investigate the amplitude frequency behavior of the system reduced order models (ROMs) are developed. Three methods are used to solve these ROMs they are 1) the method of multiple scales (MMS) for ROM with one mode of vibration, 2) homotopy analysis method (HAM) for ROM with one mode of vibration, and 3) direct numerical integration for 2 modes of vibration Reduced Order Model (2T ROM) producing time responses of the tip of the cantilever resonator. In this work the limitations of MMS and HAM are highlighted when considering large voltage values i.e hard excitations. For large voltage values MMS and HAM cannot accurately predict the amplitude frequency response; the results from 2T ROM time responses disagree significantly with the MMS and HAM solutions. The effect of voltage on the frequency response is investigated. As the voltage values in the system increase the responses shift to lower frequencies and larger amplitudes.","PeriodicalId":41412,"journal":{"name":"Mechatronic Systems and Control","volume":null,"pages":null},"PeriodicalIF":0.6,"publicationDate":"2019-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87073083","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}
Autonomous aerial manipulators have great potentials to assist humans or even fully automate manual labor-intensive tasks such as aerial cleaning, aerial transportation, infrastructure repair, and agricultural inspection and sampling. Reinforcement learning holds the promise of enabling persistent autonomy of aerial manipulators because it can adapt to different situations by automatically learning optimal policies from the interactions between the aerial manipulator and environments. However, the learning process itself could experience failures that can practically endanger the safety of aerial manipulators and hence hinder persistent autonomy. In order to solve this problem, we propose for the aerial manipulator a self-reflective learning strategy that can smartly and safely finding optimal policies for different new situations. This self-reflective manner consists of three steps: identifying the appearance of new situations, re-seeking the optimal policy with reinforcement learning, and evaluating the termination of self-reflection. Numerical simulations demonstrate, compared with conventional learning-based autonomy, our strategy can significantly reduce failures while still can finish the given task.
{"title":"Self-Reflective Learning Strategy for Persistent Autonomy of Aerial Manipulators","authors":"Xu Zhou, Jiucai Zhang, Xiaoli Zhang","doi":"10.1115/dscc2019-9086","DOIUrl":"https://doi.org/10.1115/dscc2019-9086","url":null,"abstract":"\u0000 Autonomous aerial manipulators have great potentials to assist humans or even fully automate manual labor-intensive tasks such as aerial cleaning, aerial transportation, infrastructure repair, and agricultural inspection and sampling. Reinforcement learning holds the promise of enabling persistent autonomy of aerial manipulators because it can adapt to different situations by automatically learning optimal policies from the interactions between the aerial manipulator and environments. However, the learning process itself could experience failures that can practically endanger the safety of aerial manipulators and hence hinder persistent autonomy. In order to solve this problem, we propose for the aerial manipulator a self-reflective learning strategy that can smartly and safely finding optimal policies for different new situations. This self-reflective manner consists of three steps: identifying the appearance of new situations, re-seeking the optimal policy with reinforcement learning, and evaluating the termination of self-reflection. Numerical simulations demonstrate, compared with conventional learning-based autonomy, our strategy can significantly reduce failures while still can finish the given task.","PeriodicalId":41412,"journal":{"name":"Mechatronic Systems and Control","volume":null,"pages":null},"PeriodicalIF":0.6,"publicationDate":"2019-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90265030","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：