東華大學圖書館 |

Fast, Optimal, and Safe Motion Planning for Bipedal Robots.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Fast, Optimal, and Safe Motion Planning for Bipedal Robots./
作者:	Zhao, Pengcheng.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2020,
面頁冊數:	179 p.
附註:	Source: Dissertations Abstracts International, Volume: 82-07, Section: B.
Contained By:	Dissertations Abstracts International82-07B.
標題:	Remote sensing. -
電子資源:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28240377
ISBN:	9798684626159

Fast, Optimal, and Safe Motion Planning for Bipedal Robots.
Zhao, Pengcheng.

Fast, Optimal, and Safe Motion Planning for Bipedal Robots. - Ann Arbor : ProQuest Dissertations & Theses, 2020 - 179 p.

Source: Dissertations Abstracts International, Volume: 82-07, Section: B.

Thesis (Ph.D.)--University of Michigan, 2020.

This item must not be sold to any third party vendors.

Bipedal robots have the potential to traverse a wide range of unstructured environments, which are otherwise inaccessible to wheeled vehicles.Though roboticists have successfully constructed controllers for bipedal robots to walk over uneven terrain such as snow, sand, or even stairs, it has remained challenging to synthesize such controllers in an online fashion while guaranteeing their satisfactory performance.This is primarily due to the lack of numerical method that can accommodate the non-smooth dynamics, high degrees of freedom, and underactuation that characterize bipedal robots.This dissertation proposes and implements a family of numerical methods that begin to address these three challenges along three dimensions: optimality, safety, and computational speed.First, this dissertation develops a convex relaxation-based approach to solve optimal control for hybrid systems without a priori knowledge of the optimal sequence of transition.This is accomplished by formulating the problem in the space of relaxed controls, which gives rise to a linear program whose solution is proven to compute the globally optimal controller. This conceptual program is solved using a sequence of semidefinite programs whose solutions are proven to converge from below to the true optimal solution of the original optimal control problem.Moreover, a method to synthesize the optimal controller is developed. Using an array of examples, the performance of this method is validated on problems with known solutions and also compared to a commercial solver.Second, this dissertation constructs a method to generate safety-preserving controllers for a planar bipedal robot walking on flat ground by performing reachability analysis on simplified models under the assumption that the difference between the two models can be bounded.Subsequently, this dissertation describes how this reachable set can be incorporated into a Model Predictive Control framework to select controllers that result in safe walking on the biped in an online fashion.This method is validated on a 5-link planar model.Third, this dissertation proposes a novel parallel algorithm capable of finding guaranteed optimal solutions to polynomial optimization problems up to pre-specified tolerances.Formal proofs of bounds on the time and memory usage of such method are also given.Such algorithm is implemented in parallel on GPUs and compared against state-of-the-art solvers on a group of benchmark examples.An application of such method on a real-time trajectory-planning task of a mobile robot is also demonstrated.Fourth, this dissertation constructs an online Model Predictive Control framework that guarantees safety of a 3D bipedal robot walking in a forest of randomly-placed obstacles.Using numerical integration and interval arithmetic techniques, approximations to trajectories of the robot are constructed along with guaranteed bounds on the approximation error.Safety constraints are derived using these error bounds and incorporated in a Model Predictive Control framework whose feasible solutions keep the robot from falling over and from running into obstacles.To ensure that the bipedal robot is able to avoid falling for all time, a finite-time terminal constraint is added to the Model Predictive Control algorithm.The performance of this method is implemented and compared against a naive Model Predictive Control method on a biped model with 20 degrees of freedom.In summary, this dissertation presents four methods for control synthesis of bipedal robots with improvements in either optimality, safety guarantee, or computational speed.Furthermore, the performance of all proposed methods are compared with existing methods in the field.

ISBN: 9798684626159Subjects--Topical Terms:

535394
Remote sensing.
Subjects--Index Terms:

Bipedal robot

Fast, Optimal, and Safe Motion Planning for Bipedal Robots.
LDR:05063nmm a2200445 4500 001 2281756
005 20210920103400.5
008 220723s2020 ||||||||||||||||| ||eng d
020 $a 9798684626159
035 $a (MiAaPQ)AAI28240377
035 $a (MiAaPQ)umichrackham003312
035 $a AAI28240377
040 $a MiAaPQ $c MiAaPQ
100 1 $a Zhao, Pengcheng. $3 3560464
245 1 0 $a Fast, Optimal, and Safe Motion Planning for Bipedal Robots.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2020
300 $a 179 p.
500 $a Source: Dissertations Abstracts International, Volume: 82-07, Section: B.
500 $a Advisor: Vasudevan, Ram.
502 $a Thesis (Ph.D.)--University of Michigan, 2020.
506 $a This item must not be sold to any third party vendors.
506 $a This item must not be added to any third party search indexes.
520 $a Bipedal robots have the potential to traverse a wide range of unstructured environments, which are otherwise inaccessible to wheeled vehicles.Though roboticists have successfully constructed controllers for bipedal robots to walk over uneven terrain such as snow, sand, or even stairs, it has remained challenging to synthesize such controllers in an online fashion while guaranteeing their satisfactory performance.This is primarily due to the lack of numerical method that can accommodate the non-smooth dynamics, high degrees of freedom, and underactuation that characterize bipedal robots.This dissertation proposes and implements a family of numerical methods that begin to address these three challenges along three dimensions: optimality, safety, and computational speed.First, this dissertation develops a convex relaxation-based approach to solve optimal control for hybrid systems without a priori knowledge of the optimal sequence of transition.This is accomplished by formulating the problem in the space of relaxed controls, which gives rise to a linear program whose solution is proven to compute the globally optimal controller. This conceptual program is solved using a sequence of semidefinite programs whose solutions are proven to converge from below to the true optimal solution of the original optimal control problem.Moreover, a method to synthesize the optimal controller is developed. Using an array of examples, the performance of this method is validated on problems with known solutions and also compared to a commercial solver.Second, this dissertation constructs a method to generate safety-preserving controllers for a planar bipedal robot walking on flat ground by performing reachability analysis on simplified models under the assumption that the difference between the two models can be bounded.Subsequently, this dissertation describes how this reachable set can be incorporated into a Model Predictive Control framework to select controllers that result in safe walking on the biped in an online fashion.This method is validated on a 5-link planar model.Third, this dissertation proposes a novel parallel algorithm capable of finding guaranteed optimal solutions to polynomial optimization problems up to pre-specified tolerances.Formal proofs of bounds on the time and memory usage of such method are also given.Such algorithm is implemented in parallel on GPUs and compared against state-of-the-art solvers on a group of benchmark examples.An application of such method on a real-time trajectory-planning task of a mobile robot is also demonstrated.Fourth, this dissertation constructs an online Model Predictive Control framework that guarantees safety of a 3D bipedal robot walking in a forest of randomly-placed obstacles.Using numerical integration and interval arithmetic techniques, approximations to trajectories of the robot are constructed along with guaranteed bounds on the approximation error.Safety constraints are derived using these error bounds and incorporated in a Model Predictive Control framework whose feasible solutions keep the robot from falling over and from running into obstacles.To ensure that the bipedal robot is able to avoid falling for all time, a finite-time terminal constraint is added to the Model Predictive Control algorithm.The performance of this method is implemented and compared against a naive Model Predictive Control method on a biped model with 20 degrees of freedom.In summary, this dissertation presents four methods for control synthesis of bipedal robots with improvements in either optimality, safety guarantee, or computational speed.Furthermore, the performance of all proposed methods are compared with existing methods in the field.
590 $a School code: 0127.
650 4 $a Remote sensing. $3 535394
650 4 $a Computer science. $3 523869
650 4 $a Artificial intelligence. $3 516317
650 4 $a Transportation. $3 555912
650 4 $a Web studies. $3 2122754
650 4 $a Systems science. $3 3168411
650 4 $a Mechanical engineering. $3 649730
650 4 $a Robotics. $3 519753
653 $a Bipedal robot
653 $a Optimal control
653 $a Hybrid system
653 $a Wheeled vehicles
690 $a 0771
690 $a 0790
690 $a 0984
690 $a 0646
690 $a 0800
690 $a 0548
690 $a 0799
690 $a 0709
710 2 $a University of Michigan. $b Mechanical Engineering. $3 2104815
773 0 $t Dissertations Abstracts International $g 82-07B.
790 $a 0127
791 $a Ph.D.
792 $a 2020
793 $a English
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28240377