Optimizing Long-Distance Migration in Turbulent Environments: A Breakthrough in Energy-Efficient Trajectories for Self-Propelling Agents

Significance 

Flight has always captivated the human imagination, and understanding how birds effortlessly soar over vast distances without expending excessive energy has been a subject of interest. Recent advancements in the field of autonomous systems, particularly unmanned aerial vehicles (UAVs), have prompted researchers to explore energy-efficient trajectories for long-distance migration. In a new research paper published in the peer-reviewed journal Physical Review Fluids, Ao Xu, Hua-Lin Wu, and Heng-Dong Xi from Northwestern Polytechnical University presented their innovative research on training self-propelling agents to migrate in a thermal turbulent environment using reinforcement learning algorithms.

To investigate the energy-saving capabilities of self-propelling agents in a turbulent environment, the authors employed the Rayleigh-Bénard turbulent convection cell as the training environment. The system represented the convective layer of the atmosphere and its characteristic flow patterns. By leveraging the kinetic energy present in thermal turbulence, the self-propelling agents were able to optimize their trajectories and significantly reduce energy consumption compared to naive agents planning it path in a straight line. While previous studies focused on small-aspect-ratio convection cells, migration often occurs in large-aspect-ratio systems, spanning long distances. Recognizing this, the researchers aimed to train self-propelling agents to migrate in convection cells with multiple circulation rolls. The results demonstrated the adaptability of the optimized policy obtained from the aspect ratio of Γ = 2 cell to larger Γ cells. Despite the increased complexity of flow structures and the presence of less stable horizontally stacked rolls, the smart agents successfully utilized carrier flow currents to save propelling energy. Moreover, the ratio of energy consumed by the smart agent compared to the naive agent decreased with increasing Γ, indicating even greater energy savings in larger cells.

To evaluate the robustness of the learning framework, the research team tested the optimized policy by releasing the agents from randomly chosen origins within the convection cell. They found that the success rate of migration increased with larger aspect ratios, despite the increased complexity of flow structures. The robustness of the trained agents in navigating turbulent convective environments holds promise for real-world applications, such as UAVs patrolling in the atmosphere. By adopting energy-efficient trajectories, these autonomous systems can significantly enhance their endurance and cover a wider range.

According to the authors, the research work opens up new avenues for understanding and optimizing long-distance migration in turbulent environments. The ability of self-propelling agents to explore and exploit carrier flow currents and save energy has significant implications for various domains, including unmanned aerial vehicles, gliders, and other autonomous systems operating in convective layers of the atmosphere.

As research continues to advance, we can expect to witness the implementation of energy-efficient trajectories in unmanned aerial vehicles and other autonomous platforms, revolutionizing their endurance and performance in turbulent convective environments.

Optimizing Long-Distance Migration in Turbulent Environments: A Breakthrough in Energy-Efficient Trajectories for Self-Propelling Agents - Advances in Engineering

Reference

Ao Xu , Hua-Lin Wu , and Heng-Dong Xi. Long-distance migration with minimal energy consumption in a thermal turbulent environment. Physical Review Fluids 8, 023502 (2023)

Go To Physical Review Fluids

Check Also

Advanced CSF Model for Accurate Surface Tension and Wetting Simulation in Smoothed Particle Hydrodynamics - Advances in Engineering

Advanced CSF Model for Accurate Surface Tension and Wetting Simulation in Smoothed Particle Hydrodynamics