Search papers, labs, and topics across Lattice.
This research introduces a method called LEAP, which enables soft-actuated systems to adapt to catastrophic damage in under one minute through proprioceptive feedback. By leveraging low-dimensional latent representations of damage and a robust ensemble approach, the system can effectively handle various types of damage, such as cuts and burns, in real-time without the need for extensive simulations. The findings reveal that the sample complexity for learning representations can shift from exponential to linear, significantly enhancing the adaptability of architected materials compared to traditional rigid components.
Real-time adaptation to catastrophic damage in soft robots is now achievable in under a minute, transforming their operational resilience.
Adaptation to damages and in-situ physical repairs is essential for long-term robot autonomy, yet challenging outside of narrowly defined and well-anticipated bounds. In this work we proprioceptively adapt to catastrophic damage in soft-actuated systems in under one minute. Architected materials are well equipped for adaptation: actuator failure occurs gradually rather than acutely, and damage can be described in a low-dimensional, discrete coordinate space. Surprisingly, latent damage representations plus a simple yet robust ensemble method is sufficient for adapting to unseen damage in real-time. Moreover, we identify conditions under which exponential sample complexity collapses to linear sample complexity for learned representations of architected materials, a concrete advantage over rigid components or continuum soft mechanisms. We demonstrate LEAP, our method for adaptive proprioception, via a tracing task for a 6DoF soft wrist based on Handed Shearing Auxetic (HSA) actuators. Our algorithm is able to adapt to cuts, burns, and actuator repairs, enabling simulation-free real-time adaptation that is critical for realizing the promise of soft robots outside the lab. Videos and more information are available at https://murpheylab.github.io/leap.
