Physical Artificial Intelligence in Untethered Soft Aerial Robots for Safe Environmental Interaction and Perching

As the next generation of aerial robots looks to widen its functionality and adaptability to multiple environments, the conditions required for its successful deployment has also increased. Previously, classical challenges aerial robots have had to face ranged from propeller noise, payload capacity, operational longevity, and robustness to collision and disturbances (physical obstacles and natural disturbances), just to name a few. But in order for the aerial system to effectively study natural habitats, aerial robots need to be capable of safely adapting and interacting with the unconstrained environment. One example of this is highlighted in the struggle aerial robotic systems have to tolerate collisions in order to safely land from a fall or successfully dynamically perch on objects with unknown shapes, sizes and textures.<br/><br/>To approach this problem, we made efforts to introduce compliance through the utilization of soft structures, in order to enhance mechanical impact protection, share functionality between flight and perching through shape reconfiguration, without any additional cost of reduced agility and flight time due to any extra payload.<br/>In order to accomplish this, we looked at two approaches: (1) Shared functionality between flying and perching through metamorphic bistable origami-design soft arms. (2) Utilize a lightweight, fabric, inflatable, soft-bodied structure that can pneumatically vary its body stiffness to achieve intrinsic collision resilience. This is utilized in tandem with a lightweight bistable grasper that reacts upon contact with the perching object.<br/><br/>In the first approach, we noted how gliding aerial mammals morph their arms when transitioning between gliding flight and perching. Here, the developed quadrotor arms transition between two morphing stiffness states of flying and whole-body perching, utilizing a single-direction tendon drive. The quadrotor arms are bistable and are capable of collapsing around the perch within 0.97s. The arms are developed utilizing a layered technique between carbon-fiber, polyamide, and pre-stretched latex layers. With this setup, we were able to achieve approximately 30% overall mass reduction by this shared functionality. This setup was tested both indoors and outdoors in a forest environment.<br/><br/>In the second approach, we focused on the adaptable body of soft aerial robots. We developed an inflatable woven fabric soft body and arms for the quadrotor that can intrinsically vary its body stiffness on the fly for collision resilience. With this system, we highlighted its capabilities to repeatedly endure and recover from collisions in three-dimensions, supporting even harsh falls towards the ground without breaking its structure. In tandem with an equipped hybrid fabric-based bistable grasper that curls upon impact to conform around the perching object, we demonstrated the importance of having a soft body for improving the perching success rate on different objects significantly compared to the same system with a conventional rigid quadcopter frame.<br/><br/>Overall, in order to push the boundaries of current aerial robot designs to operate in cluttered environments and extend their range of functionalities, it is important to synthesize life-like soft aerial robots with shared physical artificial intelligent features and functionality, exploiting compliance as a key to do so. This will then push the next generation of aerial robots to not only improve their aerodynamic adaptability, efficiency, or maneuverability, but also lead to novel adaptations in adaptive manipulation and perching, aerial-aquatic-terrestrial transitions, withstanding dynamic collisions, reconfiguration for traversing through tight spaces, self-healing from damages, and adhesion in multi-modal environments.