Automated shapeshifting for function recovery in damaged robots (1905.09264v1)

Abstract: A robot's mechanical parts routinely wear out from normal functioning and can be lost to injury. For autonomous robots operating in isolated or hostile environments, repair from a human operator is often not possible. Thus, much work has sought to automate damage recovery in robots. However, every case reported in the literature to date has accepted the damaged mechanical structure as fixed, and focused on learning new ways to control it. Here we show for the first time a robot that automatically recovers from unexpected damage by deforming its resting mechanical structure without changing its control policy. We found that, especially in the case of "deep insult", such as removal of all four of the robot's legs, the damaged machine evolves shape changes that not only recover the original level of function (locomotion) as before, but can in fact surpass the original level of performance (speed). This suggests that shape change, instead of control readaptation, may be a better method to recover function after damage in some cases.

• The paper demonstrates that morphological adaptation via automated shapeshifting can restore and even enhance mobility in damaged soft robots.
• It employs a silicone voxel-based quadrupedal robot to compare nine damage scenarios, revealing that shape deformation often outperforms controller-only reoptimization.
• The study underlines the potential for autonomous self-repair in challenging environments by integrating flexible materials with advanced learning techniques.

Automated Shapeshifting for Function Recovery in Damaged Robots

The paper "Automated Shapeshifting for Function Recovery in Damaged Robots" presents a novel framework for enabling robotic systems, particularly soft-bodied robots, to recover function after experiencing structural damage. While traditional approaches to robot damage recovery have focused predominantly on controller adaptation, this research posits that morphological adaptation—specifically, shape deformation—can be a superior strategy under certain damage scenarios.

Key Contributions

1. Morphological Flexibility: The research demonstrates that by altering the physical shape of a damaged robot, it's possible to restore or enhance its functionality. This goes beyond the conventional model of controller reoptimization, which adapts a fixed damaged structure to regain performance.
2. Experimental Setup: The authors construct an isobilaterally symmetrical quadrupedal robot from silicone "voxels" that can be pneumatically inflated to alter shape. This design takes advantage of recent advancements in soft robotics, wherein flexible materials enable a plethora of deformation possibilities that were previously unattainable with rigid components.
3. Comparative Analysis: A set of nine different amputation scenarios was explored to evaluate the capabilities of controller reoptimization versus morphological adaptation. Each scenario varies the extent of damage, up to the removal of three-quarters of the robot's structure or all of its legs.
4. Performance Metrics: The findings underline that the approach of automated shapeshifting outperforms controller-only adaptation in most cases concerning mobility recovery. Intriguingly, shape adaptation not only restores but in some configurations even surpasses the original, undamaged locomotion performance.

Implications

The implications of this work are multifaceted. Practically, the development of robust, self-healing robotic systems capable of operating in remote or hazardous environments, which preclude human intervention, is significantly advanced. Theoretically, it challenges prevailing paradigms in robotic resilience by demonstrating that the plasticity of a robot's physical form can compete with, or supplement, the adaptability of its control software in preserving functionality.

Future Prospects

There are several avenues for development and future exploration. Integrating machine learning models, specifically reinforcement learning in conjunction with morphological adaptation, could shorten the recovery time window by predicting optimal shape configurations post-damage. Moreover, enhancing the level of simulation fidelity with tools like multi-resolution modeling might bridge the gap further between digital model adaptations and physical implementations, a critical barrier in current sim2real challenges.

Concluding Remarks

This research revisits foundational biological insights regarding organismal adaptation and regeneration, drawing parallels to current advancements in materials science for robotics. As capability in material deformation broadens with continued research, the potential for robots capable of higher degrees of self-repair and restructuring grows. The implications extend to a wide range of applications, including lunar exploration, oceanic monitoring, and beyond, where human-led repair is unfeasible, emphasizing the strategic importance of this work in advancing autonomous, resilient systems.