Reinforcement learning-based saturated adaptive robust neural-network control of underactuated autonomous underwater vehicles

Elhaki, Omid; Shojaei, Khoshnam; Mehrmohammadi, Parisa

doi:10.1016/j.eswa.2022.116714

Cited by 41 publications

(14 citation statements)

References 57 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The proposed collision evasion controls can be integrated with trajectory tracking controls in the literature [ 32 , 33 , 34 ]. We derive the position and orientation of the AUV

at each sampling-index k [ 34 ].…”

Section: Discussionmentioning

confidence: 99%

“…Then, the straight line segment from the current AUV’s position

to the next waypoint

is set as the reference trajectory from sampling-index k to

. Trajectory tracking controls in [ 31 , 32 , 33 , 34 ] can be applied to make the AUV track the straight-line segment trajectory, even in environments with model dynamic uncertainties and the presence of external disturbances representing ocean currents and waves. Note that detailed kinematic model in AUV’s local coordinate frame is handled in the trajectory tracking controls in [ 31 , 32 , 33 , 34 ].…”

Section: Definitions and Assumptionsmentioning

confidence: 99%

See 1 more Smart Citation

Reactive Control for Collision Evasion with Extended Obstacles

Kim

2022

Sensors

View full text Add to dashboard Cite

Evading collisions in three-dimensional underwater environments is critical in exploration of an Autonomous Underwater Vehicle (AUV). In underwater environments, AUV measures an obstacle surface by utilizing a three-dimensional active sonar. This article addresses reactive collision evasion control by considering extended obstacles. Here, an extended obstacle is an arbitrary obstacle that can generate any number of measurements and not a point target generating at most one measurement. Considering 3D environments, our manuscript considers collision evasion with both moving obstacles and static obstacles. The proposed reactive collision evasion controllers are developed by considering hardware limits, such as the maximum speed or acceleration limit of an AUV. We further address how to make an AUV move towards a goal, while avoiding collision with extended obstacles. As far as we know, the proposed collision evasion controllers are novel in handling collision avoidance with an extended obstacle, in the case where an AUV measures 3D-obstacle boundaries by utilizing sonar sensors. The effectiveness of the proposed controllers is demonstrated by MATLAB simulations.

show abstract

“…The proposed collision evasion controls can be integrated with trajectory tracking controls in the literature [ 32 , 33 , 34 ]. We derive the position and orientation of the AUV

at each sampling-index k [ 34 ].…”

Section: Discussionmentioning

confidence: 99%

“…Then, the straight line segment from the current AUV’s position

to the next waypoint

is set as the reference trajectory from sampling-index k to

Section: Definitions and Assumptionsmentioning

confidence: 99%

Reactive Control for Collision Evasion with Extended Obstacles

Kim

2022

Sensors

View full text Add to dashboard Cite

show abstract

“…However, these methods require the robot’s dynamics to be identified and modeled. The algorithms in ( Elhaki et al, 2022 ; Elhaki and Shojaei, 2022 ) are designed to control multibody systems, such as tractors and underwater vehicles, without requiring detailed system models. Different from these systems, industrial robots have own control systems.…”

Section: Related Workmentioning

confidence: 99%

A data-driven approach for motion planning of industrial robots controlled by high-level motion commands

et al. 2023

View full text Add to dashboard Cite

Most motion planners generate trajectories as low-level control inputs, such as joint torque or interpolation of joint angles, which cannot be deployed directly in most industrial robot control systems. Some industrial robot systems provide interfaces to execute planned trajectories by an additional control loop with low-level control inputs. However, there is a geometric and temporal deviation between the executed and the planned motions due to the inaccurate estimation of the inaccessible robot dynamic behavior and controller parameters in the planning phase. This deviation can lead to collisions or dangerous situations, especially in heavy-duty industrial robot applications where high-speed and long-distance motions are widely used. When deploying the planned robot motion, the actual robot motion needs to be iteratively checked and adjusted to avoid collisions caused by the deviation between the planned and the executed motions. This process takes a lot of time and engineering effort. Therefore, the state-of-the-art methods no longer meet the needs of today’s agile manufacturing for robotic systems that should rapidly plan and deploy new robot motions for different tasks. We present a data-driven motion planning approach using a neural network structure to simultaneously learn high-level motion commands and robot dynamics from acquired realistic collision-free trajectories. The trained neural network can generate trajectory in the form of high-level commands, such as Point-to-Point and Linear motion commands, which can be executed directly by the robot control system. The result carried out in various experimental scenarios has shown that the geometric and temporal deviation between the executed and the planned motions by the proposed approach has been significantly reduced, even if without access to the “black box” parameters of the robot. Furthermore, the proposed approach can generate new collision-free trajectories up to 10 times faster than benchmark motion planners.

show abstract

“…Step 6. According to the new histogram, use the following mapping function to calculate the new pixel value, as shown in Equation (7). 𝐹 𝑛𝑒𝑤 and 𝐹 𝑜𝑙𝑑 is the pixel value at the midpoint (𝑥, 𝑦) of (𝑖, 𝑗), (𝑖, 𝑗 − 1), (𝑖 − 1, 𝑗), (𝑖 − 1, 𝑗 − 1) regions before and after the corresponding histogram of the image is redistributed, e_ vb, e_ vt, e_ hr, e_ Hl is the distance from the pixel at the point (𝑥, 𝑦) to the pixel point (𝑖, 𝑗), (𝑖, 𝑗 − 1), (𝑖 − 1, 𝑗), (𝑖 − 1, 𝑗 − 1).…”

Section: 𝛼 = 𝑀 𝑁mentioning

confidence: 99%

“…At present, Many enterprises are investing in the research and development of AI devices. In the field of offshore aquaculture, flexibility, accuracy, long operation time and wide working range are needed [7,8]. In offshore aquaculture, due to the influence of hydrometeorology, artificial fishing is often suspended in autumn and winter, so the whole industrial chain is stopped.…”

Section: Introductionmentioning

confidence: 99%

Sea Cucumber Identification Fishing System Based on Hog Super Pixel Feature Extraction

Wang,

Liu,

Chen

et al. 2023

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

Due to the refraction, reflection and absorption of light induced by seawater and suspended particles, the image recognition of underwater vehicle has the problem of low accuracy and difficult localization, and to solve this one problem, an improved LSTM classification algorithm (CH-LSTM) based on CLAHE and HOG is given. First, the improved CLAHE algorithm is used to enhance image by image noise removal and de-fogging, the problems of underwater image distortion, fine lines, and abrupt changes are solved in the aquaculture waters image. Then, the Histogram of Orientation Gradient (HOG) operator is used for feature extraction to describe the shape of sea cucumbers and generate feature vectors; finally, the LSTM algorithm is used to classify the feature vectors and avoid overfitting through memory gates, so the generalization ability of the model was enhanced. The identification and sorting experiments of underwater fishing sea cucumbers show that the proposed algorithm is superior, especially in the environment of poor water quality, and the identification and localization accuracy of sea cucumbers is improved compared with the traditional ground SVM and BP neural network algorithms, and the identification accuracy is above 95.28%.

show abstract

Reinforcement learning-based saturated adaptive robust neural-network control of underactuated autonomous underwater vehicles

Cited by 41 publications

References 57 publications

Reactive Control for Collision Evasion with Extended Obstacles

Reactive Control for Collision Evasion with Extended Obstacles

A data-driven approach for motion planning of industrial robots controlled by high-level motion commands

Sea Cucumber Identification Fishing System Based on Hog Super Pixel Feature Extraction

Contact Info

Product

Resources

About