2019 Fiscal Year Research-status Report
Computational research of biomimetic flight in unsteady environments with the aid of machine learning
| Project/Area Number |
18K13693
|
|---|
| Research Institution | Japan Agency for Marine-Earth Science and Technology |
|---|
Principal Investigator |
Kolomensk Dmitry 国立研究開発法人海洋研究開発機構, 付加価値情報創生部門(地球情報基盤センター), 特任研究員 (00813924)
|
|---|
| Project Period (FY) |
2018-04-01 – 2021-03-31
|
| Keywords | insect flight / bio-inspired flight / HPC |
|---|
| Outline of Annual Research Achievements |
A quasi-steady model has been developed that provides quick evaluation of the fluid-dynamic forces. This is critical for time effectiveness of the reinforcement learning process. Linear controller models have been proposed. A manuscript has been submitted on this topic (Cai et al. 2020, J. Fluid Mech., under review). Tests of reinforcement learning are in progress.
Direct import of geometrical models based on micro-CT scans into the CFD software has been implemented for steady conditions. In addition, the mass-spring model of wing deformation has undergone first evaluation using CFD coupled solver (Truong et al. 2020, Comput. Fluids 200:104426), and it is presently at the parameter identification stage. Numerical simulations of micro-insects with bristled wings have been carried out.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
Overall progress is as anticipated. Availability of new biological data on featherwing beetles has motivated inclusion of the ptiliid species in the analysis.
|
| Strategy for Future Research Activity |
The quasi-steady aerodynamic model will be combined with a simple flight dynamics model and integrated in a reinforcement learning loop. Model coefficients for featherwing beetles will be determined. Simplified phenomenological biomimetic sensorimotor models will be hypothesized. Numerical simulations of controlled free flight in perturbed air environment will be implemented. Optimal state-action policies will be obtained for each hypothesized sensorimotor model and a selected reward function. Simulated optimal trajectories in the phase plane will be compared with those observed in experiments, in order to identify biologically relevant policies. The mass-spring model of flexible wings will be fit with the mass and bending stiffness distribution from computer tomographic reconstruction and static bending experiments. The results will be used to explain the variability of wing kinematics in free flight. Conclusion on possible engineering application will be drawn.
|
| Causes of Carryover |
New numerical simulations will be performed using Tsubame supercomputer. Thus, part of the grant will be used to cover the costs of using that supercomputer facility. Also, it may become necessary to buy additional external hard disks for data storage. Since I was unable to present the results of the project at the SICB 2020 annual meeting, I plan to participate in the next meeting in January 2021, if it is held on schedule. A new manuscript containing new results has been submitted to Experiments in Fluids. If accepted, open access publication fees will be paid from the grant.
|
