Conditional Generative Adversarial Networks for Optimal Path Planning

Ma, Nachuan; Wang, Jiankun; Liu, Jianbang; Meng, Max Q.-H.

doi:10.1109/tcds.2021.3063273

Cited by 33 publications

(19 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Deep learning techniques such as Conditional Variational AutoEncoder (CVAE), CNN, GAN and their variants have been widely used to solve motion-planning algorithms by generating processed configuration space in advance to guide the expansion of classical motion-planning algorithms [19][20][21][22][23][24][25][26][27][28][29][30].…”

Section: Module Replacement Algorithmsmentioning

confidence: 99%

“…The LEGO algorithm trains the CVAE model using target samples in bottleneck regions and diverse regions to plan with sparse road maps. Takahashi et al [28] applied a GAN model with U-Net as the backbone to learning heuristic functions for grid-based A* algorithm that can reduce the search cost in 2-D space, whereas Ma et al [29] utilised the similar architecture to guide the sampling process of classical sampling-based RRT* algorithm that can accelerate the convergence speed to the optimal path significantly and generate a high-quality initial path. Figure 2 illustrates the experimental results of four maps presented in Ref.…”

Section: Cvae and Gan-related Pre-processing Modulementioning

confidence: 99%

See 1 more Smart Citation

A survey of learning‐based robot motion planning

Wang

Zhang

et al. 2021

IET Cyber-Syst and Robotics

Self Cite

View full text Add to dashboard Cite

A fundamental task in robotics is to plan collision-free motions among a set of obstacles. Recently, learning-based motion-planning methods have shown significant advantages in solving different planning problems in high-dimensional spaces and complex environments. This article serves as a survey of various different learning-based methods that have been applied to robot motion-planning problems, including supervised, unsupervised learning, and reinforcement learning. These learning-based methods either rely on a human-crafted reward function for specific tasks or learn from successful planning experiences. The classical definition and learning-related definition of motion-planning problem are provided in this article. Different learning-based motion-planning algorithms are introduced, and the combination of classical motion-planning and learning techniques is discussed in detail.This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.

show abstract

Section: Module Replacement Algorithmsmentioning

confidence: 99%

Section: Cvae and Gan-related Pre-processing Modulementioning

confidence: 99%

A survey of learning‐based robot motion planning

Wang

Zhang

et al. 2021

IET Cyber-Syst and Robotics

Self Cite

View full text Add to dashboard Cite

show abstract

“…In [27], Khan et al use Graph Neural Networks (GNNs) to encode the topology of the state space, and two different methods are proposed, one for constant graph and others for the incremental graph. [10] and [11] predict promising region to guide the sampling. However, the effects of the connectivity quality of the promising region are not stated, and the connectivity of the promising region is not enhanced.…”

Section: A Related Workmentioning

confidence: 99%

“…In [10], Neural RRT * utilizes a nonuniform sampling distribution predicted by a Convolutional Neural Network (CNN). Generative Adversarial Network (GAN) can be also implemented to achieve nonuniform sampling [11] [12]. However, they do not discuss the effect of the connectivity of the predicted promising region.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Enhance Connectivity of Promising Regions for Sampling-based Path Planning

Ma,

Li,

Liu

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Sampling-based path planning algorithms usually implement uniform sampling methods to search the state space. However, uniform sampling may lead to unnecessary exploration in many scenarios, such as the environment with a few dead ends. Our previous work proposes to use the promising region to guide the sampling process to address the issue. However, the predicted promising regions are often disconnected, which means they cannot connect the start and goal state, resulting in a lack of probabilistic completeness. This work focuses on enhancing the connectivity of predicted promising regions. Our proposed method regresses the connectivity probability of the edges in the x and y directions. In addition, it calculates the weight of the promising edges in loss to guide the neural network to pay more attention to the connectivity of the promising regions. We conduct a series of simulation experiments, and the results show that the connectivity of promising regions improves significantly. Furthermore, we analyze the effect of connectivity on samplingbased path planning algorithms and conclude that connectivity plays an essential role in maintaining algorithm performance.Note to Practitioners-This work is derived from the promising region prediction for sampling-based path planning. The sampling-based path planning methods have been widely used in robotics due to their efficiency. To further improve the efficiency of these algorithms, sampling in the promising region predicted by a neural network is introduced into the sampling procedure. However, the connectivity of the promising region has yet to be considered, and it will affect the performance of the algorithms in several aspects. To demonstrate this problem, we compare the performance of the neural heuristic algorithms under different connectivity statuses in this paper. Furthermore, to enhance the connectivity of the predicted promising region, the novel prediction output and loss function are proposed. The simulation results show improvements in the algorithms after utilizing our method.

show abstract

Artificial Intelligence Technology for Path Planning of Automated Earthwork Machinery

Zhou,

Wang,

et al. 2024

Journal of Field Robotics

View full text Add to dashboard Cite

The challenging characteristics of earthwork environments—complex, unstructured, and constantly evolving—pose significant challenges for the path planning of automated earthwork machinery. Recent advancements in artificial intelligence (AI) technology have opened new avenues to address these challenges, which are crucial for improving the intelligence level of automated earthwork machinery. However, there is a notable lack of comprehensive analyses on AI‐based path planning in earthwork operations. Consequently, we provide a systematic review of four AI technologies currently employed in path planning for earthwork machinery, including (1) evolutionary computation, (2) swarm intelligence, (3) machine learning, and (4) other AI‐based technologies. We analyzed the application and performance evaluation results of these technologies across various construction machinery. Through this systematic analysis, we identified several key challenges: (1) multiconstraint earthwork environments, (2) generalization across 3D unstructured sites, (3) adaptability to dynamically uncertain environments, and (4) shortage of on‐site validation. We then outline potential future directions: (1) integration of generative AI with reinforcement learning, (2) use of large model technology, (3) adoption of embodied intelligence technology, and (4) conduction of more on‐site experiments.

show abstract

Conditional Generative Adversarial Networks for Optimal Path Planning

Cited by 33 publications

References 22 publications

A survey of learning‐based robot motion planning

A survey of learning‐based robot motion planning

Enhance Connectivity of Promising Regions for Sampling-based Path Planning

Artificial Intelligence Technology for Path Planning of Automated Earthwork Machinery

Contact Info

Product

Resources

About