Article
Robotics
Ellis Ratner, Andrea Bajcsy, Terrence Fong, Claire J. Tomlin, Anca D. Dragan
Summary: The study introduces an adaptive model estimation algorithm for robotic systems operating under changing dynamics, efficiently estimating and adapting to changes to improve efficiency. By maintaining a small subset of models at each time step and choosing appropriate models to keep, performance is not compromised while efficiency is enhanced.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2021)
Article
Robotics
Zachary Kingston, Lydia E. Kavraki
Summary: Robotic manipulation often involves multimodal planning tasks, which require finding a sequence of transitions between modes. However, many multimodal planners fail to scale when faced with difficult motion planning or tasks with a long horizon. This work proposes a solution that uses experience-based planning and a layered planning approach to improve the scalability and task satisfaction of multimodal planners, enabling them to handle complex manipulation tasks and achieve significant improvements in high-dimensional robot scenes.
IEEE TRANSACTIONS ON ROBOTICS
(2023)
Article
Robotics
Simon Zimmermann, Roi Poranne, Stelian Coros
Summary: This letter presents model-based optimal control techniques for dynamic manipulation of deformable objects, focusing on the application of both batch Newton method and stagewise Differential Dynamic Programming (DDP) approach. The experiments and analysis show the performance of these methods depends heavily on the dimensions of the control problems, with implicit integration schemes being necessary for stable simulation. The efficacy of the trajectory optimization formulations is demonstrated through a variety of simulation and real-world experiments.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2021)
Article
Automation & Control Systems
Shuai Liu, Pengcheng Liu
Summary: This paper optimizes sampling-based motion planning algorithms to find the most suitable motion planner for different scenes and queries. STOMP performs well in low-complexity scenes, but the optimization performance in more complex scenes may not be as good as the original algorithm.
INTERNATIONAL JOURNAL OF ADVANCED MANUFACTURING TECHNOLOGY
(2022)
Article
Computer Science, Interdisciplinary Applications
Xingyi Yao, Wenhua Li, Xiaogang Pan, Rui Wang
Summary: This study focuses on the multi-objective path planning problem and proposes a new solution-encoding method and environmental selection strategy to address the multi-modal minimum path problems. The experiments prove that the proposed method is effective and efficient for multimodal multi-objective path planning.
COMPUTERS & INDUSTRIAL ENGINEERING
(2022)
Article
Engineering, Civil
Alexander Botros, Stephen L. Smith
Summary: Lattice-based planning techniques simplify motion planning for autonomous vehicles by limiting available motions to a pre-computed set of primitives. This study formulates the problem for an arbitrary lattice as a mixed integer linear program and proposes an A*-based algorithm to solve the motion planning problem using these primitives. The study also introduces an algorithm that removes excessive oscillations from planned motions. The method is validated for autonomous driving in both parking lot and highway scenarios.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
(2023)
Article
Robotics
Jonas Wittmann, Daniel J. Rixen
Summary: This study proposes a method to improve the efficiency of collaborative robot systems through cubic splines that compute minimum trajectory durations while respecting actuator limits and C-2 continuity. The research demonstrates significant progress in analyzing the gradients of the underlying nonlinear optimization problem, leading to an efficient solution approach.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2022)
Article
Robotics
Yongxing Tang, Zhanxia Zhu, Hongwen Zhang
Summary: By defining a modified time informed set (MTIS), this paper solves the time-optimal kinodynamic motion planning problem by limiting the planning domain. Compared to the original time informed set (TIS), MTIS not only saves the running time of sampling-based motion planners (SBMP), but also extends the applicable scope. Additionally, a spatio-temporal sampling strategy adapted to MTIS is proposed, which accelerates convergence and reduces memory requirement when combined with common SBMP.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2023)
Article
Robotics
Takayuki Osa
Summary: The objective function in trajectory optimization can be non-convex and have an infinite set of local optima, leading to diverse solutions for a given task. To tackle this issue, researchers propose an optimization method that learns an infinite set of solutions by obtaining diverse solutions through learning latent representations. This approach can be interpreted as training a deep generative model for generating collision-free trajectories in motion planning, demonstrating the representation of an infinite set of homotopic solutions for motion planning problems in experimental results.
INTERNATIONAL JOURNAL OF ROBOTICS RESEARCH
(2022)
Article
Computer Science, Interdisciplinary Applications
Fabio Rodriguez, Jose-Miguel Diaz-Banez, Ernesto Sanchez-Laulhe, Jesus Capitan, Anibal Ollero
Summary: This paper presents a novel algorithm for energy-efficient trajectory planning. The algorithm minimizes energy consumption by combining gliding and flapping maneuvers, and applies heuristic search and reference curves for guidance in online planning.
COMPUTERS & INDUSTRIAL ENGINEERING
(2022)
Article
Computer Science, Information Systems
Yu Sun, Guanqin Pan, Yaoshen Li, Yingying Yang
Summary: This study proposes a method that combines density clustering technique and differential evolutionary algorithm to solve multimodal optimization problems. Experimental results on CEC2013 benchmark function demonstrate the effectiveness of the proposed method.
INFORMATION SCIENCES
(2023)
Article
Robotics
Ashkan Jasour, Weiqiao Han, Brian C. Williams
Summary: In this paper, we propose a risk-bounded trajectory planning method to address the trajectory planning problem in uncertain nonconvex static and dynamic environments. By leveraging risk contours and convex methods based on sum-of-squares optimization, we can efficiently obtain continuous-time risk-bounded trajectories without time discretization.
INTERNATIONAL JOURNAL OF ROBOTICS RESEARCH
(2023)
Article
Robotics
Kris Hauser
Summary: This article introduces a novel optimization method that handles collision constraints with complex, non-convex 3D geometries using a semi-infinite program and efficient implementation of collision detection with signed distance field and point clouds. Experimental results show improved performance and efficient collision avoidance compared to optimizers dealing with convex polyhedra only in pose and trajectory optimization for free-flying rigid bodies and articulated robots.
INTERNATIONAL JOURNAL OF ROBOTICS RESEARCH
(2021)
Article
Automation & Control Systems
Zi-Jia Wang, Yu-Ren Zhou, Jun Zhang
Summary: This article presents a parameter-free niching method based on adaptive estimation distribution (AED) and develops a distributed differential evolution (DDE) algorithm, called AED-DDE, for solving multimodal optimization problems (MMOPs). The algorithm improves population diversity through a multiniche co-evolution mechanism and refines solution accuracy through probabilistic local search.
IEEE TRANSACTIONS ON CYBERNETICS
(2022)
Article
Computer Science, Information Systems
Miao Wang, Qingshan Liu, Yanling Zheng
Summary: The paper introduces a novel velocity planning method based on path curvature, which is implemented in three steps to achieve minimum moving time. By dividing the path into elementary parts and transforming the velocity planning into an unconstrained optimization problem, the proposed method uses a modified PPSO algorithm to obtain the time-optimal velocity profile. Experimental results demonstrate that the method can generate a smooth time-optimal velocity profile while avoiding sudden acceleration changes.
INFORMATION SCIENCES
(2021)
Article
Automation & Control Systems
Takayuki Osa, Satoshi Uchida, Naohiko Sugita, Mamoru Mitsuishi
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2015)
Article
Automation & Control Systems
Takayuki Osa, Christian Farid Abawi, Naohiko Sugita, Hirotaka Chikuda, Shurei Sugita, Takeyuki Tanaka, Hirofumi Oshima, Toru Moro, Sakae Tanaka, Mamoru Mitsuishi
IEEE-ASME TRANSACTIONS ON MECHATRONICS
(2015)
Article
Robotics
T. Osa, Jan Peters, G. Neumann
Article
Engineering, Biomedical
Naohiko Sugita, Takayuki Osa, Mamoru Mitsuishi
MEDICAL ENGINEERING & PHYSICS
(2009)
Article
Robotics
Takayuki Osa
Summary: The objective function in trajectory optimization can be non-convex and have an infinite set of local optima, leading to diverse solutions for a given task. To tackle this issue, researchers propose an optimization method that learns an infinite set of solutions by obtaining diverse solutions through learning latent representations. This approach can be interpreted as training a deep generative model for generating collision-free trajectories in motion planning, demonstrating the representation of an infinite set of homotopic solutions for motion planning problems in experimental results.
INTERNATIONAL JOURNAL OF ROBOTICS RESEARCH
(2022)
Article
Multidisciplinary Sciences
Kohei Nagai, Takayuki Osa, Gen Inoue, Takuya Tsujiguchi, Takuto Araki, Yoshiyuki Kuroda, Morio Tomizawa, Keisuke Nagato
Summary: In this study, a Bayesian optimization method is used to explore the production process parameters of powder film forming. By combining experimental and data-driven approaches, the optimal parameter sets were found with a significantly reduced number of experiments. The mechanism of film formation was also inferred.
SCIENTIFIC REPORTS
(2022)
Article
Computer Science, Information Systems
Takayuki Osa, Masanori Aizawa
Summary: This study investigates the use of depth images and a variant of Q-learning algorithm called Qt-Opt to learn the excavation task. A regularization method using virtual adversarial samples to reduce overestimation of Q-values is proposed. The results demonstrate that Qt-Opt is more sample-efficient than other methods and the choice of policy representation is crucial for performance.
Article
Computer Science, Artificial Intelligence
Takayuki Osa, Voot Tangkaratt, Masashi Sugiyama
Summary: Reinforcement learning algorithms are often limited to learning a single solution for a specified task, but recent studies have shown the benefits of learning diverse solutions. This study proposes a novel method that overcomes the bias problem and successfully learns diverse solutions by directly maximizing the variational lower bound of mutual information.
Article
Robotics
Hiroyuki Karasawa, Tomohiro Kanemaki, Kei Oomae, Rui Fukui, Masayuki Nakao, Takayuki Osa
IEEE ROBOTICS AND AUTOMATION LETTERS
(2020)
Proceedings Paper
Computer Science, Artificial Intelligence
Takayuki Osa, Masashi Sugiyama
THIRTY-SECOND AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE / THIRTIETH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE / EIGHTH AAAI SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE
(2018)
Proceedings Paper
Robotics
Takayuki Osa, Jan Peters, Gerhard Neumann
2016 INTERNATIONAL SYMPOSIUM ON EXPERIMENTAL ROBOTICS
(2017)
Proceedings Paper
Computer Science, Artificial Intelligence
Takayuki Osa, Satoshi Uchida, Naohiko Sugita, Mamoru Mitsuishi
2014 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS 2014)
(2014)
Proceedings Paper
Computer Science, Artificial Intelligence
Takayuki Osa, Takuto Haniu, Kanako Harada, Naohiko Sugita, Mamoru Mitsuishi
2013 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS)
(2013)
Article
Robotics
Takayuki Osa, Amir M. Ghalamzan Esfahani, Rustam Stolkin, Rudolf Lioutikov, Jan Peters, Gerhard Neumann
IEEE ROBOTICS AND AUTOMATION LETTERS
(2017)