Next Best View Planning via Reinforcement Learning for Scanning of Arbitrary 3D Shapes

Potapova, S. G.; Artemov, Alexey; Sviridov, Sergey; Musatkina, D. A.; Zorin, Denis; Burnaev, Evgeniy

doi:10.1134/s1064226920120141

Cited by 7 publications

(4 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In another approach, a sampling-based method was proposed to analyse and improve the probability of coverage completeness and convergence for a specific structural inspection [9]. Other approaches exploited chaotic model [10], artificial potential field [11], Spanning Tree Covering [12], cellular and polygon decomposition [13], Probabilistic road map [14], particle swarm [15], neural network [16], and reinforcement learning [17] to optimise different criteria for travelling time, energy consumption, trajectory smoothness, and collision risk.…”

Section: Introductionmentioning

confidence: 99%

Unified Robust Path Planning and Optimal Trajectory Generation for Efficient 3D Area Coverage of Quadrotor UAVs

Rekabi-Bana,

Hu,

Krajník

et al. 2024

IEEE Trans. Intell. Transport. Syst.

View full text Add to dashboard Cite

Area coverage is an important problem in robotics applications, which has been widely used in search and rescue, offshore industrial inspection, and smart agriculture. This paper demonstrates a novel unified robust path planning, optimal trajectory generation, and control architecture for a quadrotor coverage mission. To achieve safe navigation in uncertain working environments containing obstacles, the proposed algorithm applies a modified probabilistic roadmap to generating a connected search graph considering the risk of collision with the obstacles. Furthermore, a recursive node and link generation scheme determines a more efficient search graph without extra complexity to reduce the computational burden during the planning procedure. An optimal three-dimensional trajectory generation is then suggested to connect the optimal discrete path generated by the planning algorithm, and the robust control policy is designed based on the cascade N LH∞ framework. The integrated framework is capable of compensating for the effects of uncertainties and disturbances while accomplishing the area coverage mission. The feasibility, robustness and performance of the proposed framework are evaluated through Monte Carlo simulations, PX4 Software-In-the-Loop test facility, and realworld experiments.

show abstract

Section: Introductionmentioning

confidence: 99%

Unified Robust Path Planning and Optimal Trajectory Generation for Efficient 3D Area Coverage of Quadrotor UAVs

Rekabi-Bana,

Hu,

Krajník

et al. 2024

IEEE Trans. Intell. Transport. Syst.

View full text Add to dashboard Cite

show abstract

“…However, this constraint is relaxed in the active sensing scenario, where additional observation can be acquired to improve the quality of the 3D reconstructions. In active vision [1], for instance, the objective can be to iteratively select camera perspectives from an object that result in the highest improvement in quality of the reconstruction [71] and only very recently the research community has started to leverage large scale datasets to learn exploration strategies that generalize to unseen objects [72,2,37,45,24,73,44,3].…”

Section: Introductionmentioning

confidence: 99%

“…Similar to our setting, active vision can be useful in 3D shape reconstruction [58,65,14,37,11,32], but where camera perspectives are planned instead of grasp locations. Only very recently have deep learning active vision approaches been proposed for 3D object reconstruction [72,2,37,45]. Deep learning based active sensing for reconstruction has also recently emerged in the medical imaging domain, where the time spent performing MRI scans has been reduced by learning to select a small number of more informative frequencies over a pre-trained reconstruction model [24,73,44,3].…”

Section: Introductionmentioning

confidence: 99%

Active 3D Shape Reconstruction from Vision and Touch

Smith¹,

Meger²,

Villaseñor-Pineda³

et al. 2021

Preprint

View full text Add to dashboard Cite

Humans build 3D understandings of the world through active object exploration, using jointly their senses of vision and touch. However, in 3D shape reconstruction, most recent progress has relied on static datasets of limited sensory data such as RGB images, depth maps or haptic readings, leaving the active exploration of the shape largely unexplored. In active touch sensing for 3D reconstruction, the goal is to actively select the tactile readings that maximize the improvement in shape reconstruction accuracy. However, the development of deep learning-based active touch models is largely limited by the lack of frameworks for shape exploration. In this paper, we focus on this problem and introduce a system composed of: 1) a haptic simulator leveraging high spatial resolution vision-based tactile sensors for active touching of 3D objects; 2) a mesh-based 3D shape reconstruction model that relies on tactile or visuotactile signals; and 3) a set of data-driven solutions with either tactile or visuotactile priors to guide the shape exploration. Our framework enables the development of the first fully datadriven solutions to active touch on top of learned models for object understanding. Our experiments show the benefits of such solutions in the task of 3D shape understanding where our models consistently outperform natural baselines. We provide our framework as a tool to foster future research in this direction.

show abstract

“…[327]. Metaheuristic algorithm is superior in solving small workspace but can get stuck in local minima and the computational complexity exponentially increases when the workspace expands.…”

mentioning

confidence: 99%

A Comprehensive Review of Coverage Path Planning in Robotics Using Classical and Heuristic Algorithms

Mohd‐Mokhtar²,

2021

View full text Add to dashboard Cite

The small battery capacities of the mobile robot and the un-optimized planning efficiency of the industrial robot bottlenecked the time efficiency and productivity rate of coverage tasks in terms of speed and accuracy, putting a great constraint on the usability of the robot applications in various planning strategies in specific environmental conditions. Thus, it became highly desirable to address the optimization problems related to exploration and coverage path planning (CPP). In general, the goal of the CPP is to find an optimal coverage path with generates a collision-free trajectory by reducing the travel time, processing speed, cost energy, and the number of turns along the path length, as well as low overlapped rate, which reflect the robustness of CPP. This paper reviews the principle of CPP and discusses the development trend, including design variations and the characteristic of optimization algorithms, such as classical, heuristic, and most recent deep learning methods. Then, we compare the advantages and disadvantages of the existing CPP-based modeling in the area and target coverage. Finally, we conclude numerous open research problems of the CPP and make suggestions for future research directions to gain insights. INDEX TERMSCoverage path planning, exploration, heuristic algorithm, deep reinforcement learning.

show abstract

Next Best View Planning via Reinforcement Learning for Scanning of Arbitrary 3D Shapes

Cited by 7 publications

References 11 publications

Unified Robust Path Planning and Optimal Trajectory Generation for Efficient 3D Area Coverage of Quadrotor UAVs

Unified Robust Path Planning and Optimal Trajectory Generation for Efficient 3D Area Coverage of Quadrotor UAVs

Active 3D Shape Reconstruction from Vision and Touch

A Comprehensive Review of Coverage Path Planning in Robotics Using Classical and Heuristic Algorithms

Contact Info

Product

Resources

About