Propulsion optimization of a jellyfish-inspired robot based on a nonintrusive reduced-order model with proper orthogonal decomposition

Abstract

In this research, the propulsion of the proposed jellyfish-inspired mantle undulated propulsion robot (MUPRo) is optimized. To reliably predict the hydrodynamic forces acting on the robot, the proposed non-intrusive reduced-order model (NIROM) based on proper orthogonal decomposition (POD) additionally considers the POD basis that has an important contribution to the features on the specified boundary. The proposed model establishes a mapping between the parameter-driven motion of the mantle and the evolution of the fluid characteristics around the swimmer. Moreover, to predict new cases where the input needs to be updated, the input of the proposed model is taken from the kinematics of the robot rather than extracted from full-order high-fidelity models. In this paper, it takes approximately 950s to perform a simulation using the full-order high-fidelity model. However, the computational cost for one prediction with the proposed POD-NIROM is around 0.54s, of which about 0.2s is contributed by preprocessing. Compared with the NIROM based on the classic POD method, the proposed POD-NIROM can effectively update the input and reasonably predict the characteristics on the boundary. The analysis of the hydrodynamic performance of the MUPRo pinpoints that, under a certain period and a certain undulation amplitude, the hydrodynamic force generated by swinging-like mantle motion (k<0.5) is greater, outperforming Aequorea victoria in startup acceleration. It is demonstrated that considering a certain power loss and a certain tail beat amplitude, the wave-like mantle motion (k>0.5) can produce greater propulsion, which means higher propulsion efficiency.

A two-dimensional hydrodynamics prediction framework for mantle-undulated propulsion robot using multiple proper orthogonal decomposition and long short term memory neural network

“…Studies have revealed that the size of a computational domain affects the solution accuracy of unsteady problems [57][58][59]. Based on this, the size of the computational domain is strictly consistent with the size used in the 2D prototype experiments (figure 1(C)), the details can be find in the recent work [29]. As shown in figure 1(B), the size of the 2D numerical pool is 2000 mm × 500 mm, the mantle starts at (500 mm, 120 mm), and the front of the main cabin is located at (300 mm, 0).…”

“…Moreover, we applied axisymmetric assumptions in the HF model and the subsequent ROM to reduce the number of elements required for the simulation to converge, since both the computational domain and the undulating motion of the mantle are symmetric about the x-axis. However, it is worth mentioning that the lofting result for the 2D computational domain, where the rotational symmetry assumption is applied, does not entirely agree with the 3D prototype due to the neglect of the connection between the mantle and the main cabin [29].…”

“…The exploration and discussion of jellyfish propulsion can provide guidance in the control of propulsion and optimisation of parameters in a jellyfish-inspired robot [20]. The hydrodynamic mechanisms of jellyfish, as well as the jellyfishinspired robots, have garnered significant attention from the scientific community through experimental measurements [12,15,[21][22][23], numerical simulations [24][25][26][27][28], and reduced-order models [29].…”

“…When discussing the instantaneous hydrodynamic force acting on a jellyfish-inspired robot, it is expected that fluid features near the robot can be reliably predicted. However, the identification of fluid features in the vicinity of the robot poses a challenge if the classical POD algorithm is used because the solution states in close proximity to the boundary have a low contribution to the total system energy and tend to be ignored by POD-based reduced-order models (ROMs) [29]. To address this issue, a multiple POD (MPOD) algorithm is proposed in this paper for the identification of fluid features near the robot.…”

“…The classic POD-based ROM details are discussed in the existing literature [31,32,38,[47][48][49][50]. In recent years, non-intrusive models based on machine learning (ML) have been proposed, which have significantly enhanced the stability and computational efficiency of the POD-based ROM [29,51,52]. Furthermore, studies have demonstrated that non-intrusive ROMs (NIROMs) based on ML produce outcomes similar to those of the POD-based reconstructions [51,53,54].…”