東華大學圖書館 |

Model-Based Reinforcement Learning for Truck Trailer Robotics Vehicle's Trajectory Planning and Control /

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Model-Based Reinforcement Learning for Truck Trailer Robotics Vehicle's Trajectory Planning and Control // Hao Yan.
作者:	Yan, Hao,
面頁冊數:	1 electronic resource (188 pages)
附註:	Source: Dissertations Abstracts International, Volume: 86-07, Section: B.
Contained By:	Dissertations Abstracts International86-07B.
標題:	Robotics. -
電子資源:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31633374
ISBN:	9798302837264

Model-Based Reinforcement Learning for Truck Trailer Robotics Vehicle's Trajectory Planning and Control /
Yan, Hao,

Model-Based Reinforcement Learning for Truck Trailer Robotics Vehicle's Trajectory Planning and Control /Hao Yan. - 1 electronic resource (188 pages)

Source: Dissertations Abstracts International, Volume: 86-07, Section: B.

Advanced Driver Assistance Systems (ADAS) has become increasingly important in the automotive industry, because of customers' need for higher safety and efficiency for transportation. Truck-trailer systems, which occupies a large market share in both commercial and recreational sections, have been facing special challenges for ADAS due to the complex vehicle system dynamics and special skills required for maneuvering, especially in reverse. In the current market products, traditional control methods consisted of path planning and classical control methods have been widely used, however, they often struggle to adapt to the changing scenarios in real-world applications. On the other hand, with the development of Machine Learning, one of its research field, reinforcement learning, shows great potential in robotics and control area. The reinforcement learning algorithm is a semi supervised machine learning approach, and it can provide a flexible solution by enabling systems to learn and improve through experience, potentially conquer the limitations of traditional approaches. This dissertation explores the development of control systems for Truck-trailer Robotics Vehicle (TTRV) system, especially during reverse driving and parking in low-speed, closed, and and unstructured environments. The research addressed challenges caused by the complex dynamics of TTRV systems, which is known for its nonlinearity, nonholonomic constraints, and under-actuation behaviours. And this research applied several modified reinforcement learning approaches to trailer control, as well as traditional control methods, to show the improved ADAS capabilities of reinforcement learning approaches, and effectiveness of satisfying the growing need for intelligent transportation solutions.This dissertation begins with a review of classical control methods which has been used for truck trailer system, including pole placement, Lyapunov controller, Proportional-Integral-Derivative (PID), Linear Quadratic Regulator (LQR), Model Predictive Control (MPC), and other nonlinear control methods. Multiple path planning approaches such as Dubins paths, Reeds-Shepp paths, and the A* algorithm are also applied and tested to the articulated vehicle. While these methods have shown capability in certain scenarios, changing environments, online calibration, and complex dynamics can make traditional algorithm struggle in real application situations. Moreover, these algorithms require pre-design and fine-tuning from engineers, which limits their application and self-evolution in real-world situations. All these problems slow down the popularity for advanced TTRV control systems in nowadays fast growing market.To overcome these limitations mentioned above, an end-to-end Deep Reinforcement Learning (DRL) network solution for TTRV reverse autonomous control is proposed in this dissertation. The study compares several DRL algorithms with different computational complexities and performances, including both on-policy and off-policy approaches, all of them showed the effective of DRL in solving the trailer control problem.Moreover, a hybrid model-based reinforcement learning approach is developed in this research, which integrates classical path planning techniques into the DRL algorithm to improve sample efficiency and performance. The proposed algorithm is tested in several obstacle avoidance simulations, demonstrating its ability to handle target chasing, obstacle avoidance, and passenger comfort simultaneously. Also, an Imitation Inspired Probabilistic Ensembles with Trajectory Sampling (IIPETS) algorithm is developed on the base of Probabilistic Ensembles with Trajectory Sampling (PETS) algorithm, this model based approach significantly improved sample efficiency, and showed outstanding robustness and adaptability for controlling the truck trailer system.The results of this study contribute to the development of intelligent ADAS technologies for truck trailer systems, which can potentially satisfy the growing demand for efficient and safe transportation market needs. The developed Reinforcement Learning (RL) based controllers demonstrated the adaptation capability to changing scenarios and overcoming the limitations of classical control methods, paving the way for a more robust and flexible TTRV intelligent control systems.

English

ISBN: 9798302837264Subjects--Topical Terms:

519753
Robotics.
Subjects--Index Terms:

Path planning

Model-Based Reinforcement Learning for Truck Trailer Robotics Vehicle's Trajectory Planning and Control /
LDR:05801nmm a22004333i 4500 001 2437691
005 20251124054958.5
006 m o d
007 cr|nu||||||||
008 261011s2024 miu||||||m |||||||eng d
020 $a 9798302837264
035 $a (MiAaPQD)AAI31633374
035 $a AAI31633374
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Yan, Hao, $e author. $3 3828762
245 1 0 $a Model-Based Reinforcement Learning for Truck Trailer Robotics Vehicle's Trajectory Planning and Control / $c Hao Yan.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2024
300 $a 1 electronic resource (188 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 86-07, Section: B.
500 $a Advisors: Zohdy, Mohamed Committee members: Kaur, Amanpreet; Li, Li; Monroe, Ryan.
502 $b Ph.D. $c Oakland University $d 2024.
520 $a Advanced Driver Assistance Systems (ADAS) has become increasingly important in the automotive industry, because of customers' need for higher safety and efficiency for transportation. Truck-trailer systems, which occupies a large market share in both commercial and recreational sections, have been facing special challenges for ADAS due to the complex vehicle system dynamics and special skills required for maneuvering, especially in reverse. In the current market products, traditional control methods consisted of path planning and classical control methods have been widely used, however, they often struggle to adapt to the changing scenarios in real-world applications. On the other hand, with the development of Machine Learning, one of its research field, reinforcement learning, shows great potential in robotics and control area. The reinforcement learning algorithm is a semi supervised machine learning approach, and it can provide a flexible solution by enabling systems to learn and improve through experience, potentially conquer the limitations of traditional approaches. This dissertation explores the development of control systems for Truck-trailer Robotics Vehicle (TTRV) system, especially during reverse driving and parking in low-speed, closed, and and unstructured environments. The research addressed challenges caused by the complex dynamics of TTRV systems, which is known for its nonlinearity, nonholonomic constraints, and under-actuation behaviours. And this research applied several modified reinforcement learning approaches to trailer control, as well as traditional control methods, to show the improved ADAS capabilities of reinforcement learning approaches, and effectiveness of satisfying the growing need for intelligent transportation solutions.This dissertation begins with a review of classical control methods which has been used for truck trailer system, including pole placement, Lyapunov controller, Proportional-Integral-Derivative (PID), Linear Quadratic Regulator (LQR), Model Predictive Control (MPC), and other nonlinear control methods. Multiple path planning approaches such as Dubins paths, Reeds-Shepp paths, and the A* algorithm are also applied and tested to the articulated vehicle. While these methods have shown capability in certain scenarios, changing environments, online calibration, and complex dynamics can make traditional algorithm struggle in real application situations. Moreover, these algorithms require pre-design and fine-tuning from engineers, which limits their application and self-evolution in real-world situations. All these problems slow down the popularity for advanced TTRV control systems in nowadays fast growing market.To overcome these limitations mentioned above, an end-to-end Deep Reinforcement Learning (DRL) network solution for TTRV reverse autonomous control is proposed in this dissertation. The study compares several DRL algorithms with different computational complexities and performances, including both on-policy and off-policy approaches, all of them showed the effective of DRL in solving the trailer control problem.Moreover, a hybrid model-based reinforcement learning approach is developed in this research, which integrates classical path planning techniques into the DRL algorithm to improve sample efficiency and performance. The proposed algorithm is tested in several obstacle avoidance simulations, demonstrating its ability to handle target chasing, obstacle avoidance, and passenger comfort simultaneously. Also, an Imitation Inspired Probabilistic Ensembles with Trajectory Sampling (IIPETS) algorithm is developed on the base of Probabilistic Ensembles with Trajectory Sampling (PETS) algorithm, this model based approach significantly improved sample efficiency, and showed outstanding robustness and adaptability for controlling the truck trailer system.The results of this study contribute to the development of intelligent ADAS technologies for truck trailer systems, which can potentially satisfy the growing demand for efficient and safe transportation market needs. The developed Reinforcement Learning (RL) based controllers demonstrated the adaptation capability to changing scenarios and overcoming the limitations of classical control methods, paving the way for a more robust and flexible TTRV intelligent control systems.
546 $a English
590 $a School code: 0446
650 4 $a Robotics. $3 519753
650 4 $a Automotive engineering. $3 2181195
653 $a Path planning
653 $a Reinforcement learning
653 $a Truck trailer
653 $a Vehicle control
653 $a Automotive industry
690 $a 0800
690 $a 0771
690 $a 0540
710 2 $a Oakland University. $b Electrical and Computer Engineering. $e degree granting institution. $3 3819703
720 1 $a Zohdy, Mohamed $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 86-07B.
790 $a 0446
791 $a Ph.D.
792 $a 2024
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31633374