Bio-inspired compensatory strategies for damage to flapping robotic propulsors

Area of Science:

• Robotics
• Bio-inspired Engineering
• Machine Learning

Background:

• Natural swimmers and flyers exhibit remarkable resilience to propulsor damage, altering mechanics to maintain function.
• Directly applying biological damage recovery strategies to robotic systems may be suboptimal due to differing evolutionary pressures.

Purpose of the Study:

• To investigate optimal autonomous repair strategies for robotic flapping propulsors using machine learning.
• To compare damage recovery adaptations in robotic systems with those observed in natural organisms.

Main Methods:

• Implemented an online artificial evolution system with hardware-in-the-loop testing.
• Utilized a flexible plate as the robotic flapping propulsor.
• Employed machine learning to compare biological and robotic damage recovery adaptations.

Main Results:

• For thrust recovery, the learned robotic strategy significantly increased amplitude, frequency, and angle of attack (AOA), including phase-shifting AOA by ~110°.
• Side force recovery correlated with AOA, with no clear amplitude or frequency trends found in the robotic system.
• Observed robotic adaptations differed from those typically reported in fish and insect literature for damage recovery.

Conclusions:

• Optimal mechanical flapping propulsor adaptations for damage repair may not align with natural biological strategies.
• Machine learning can identify unique, efficient repair mechanisms for robotic systems.
• This research advances bio-inspired robotics and autonomous systems capable of self-repair.