We study online learning in adversarial nonstationary environments. Since the future can be very different from the past, a critical challenge is to gracefully forget the history while new data comes in. To formalize this intuition, we revisit the discounted regret in online convex optimization, and propose an adaptive (i.e., instance optimal), FTRL-based algorithm that improves the widespread non-adaptive baseline – gradient descent with a constant learning rate. From a practical perspective, this refines the classical idea of regularization in lifelong learning: we show that designing better regularizers can be guided by the principled theory of adaptive online optimization. Complementing this result, we also consider the (Gibbs & Candes, 2021)-style online conformal prediction problem, where the goal is to sequentially predict the uncertainty sets of a black-box machine learning model. We show that the FTRL nature of our algorithm can simplify the conventional gradient-descent-based analysis, leading to instance-dependent performance guarantees.

Realizing high-performance soft robotic grippers is challenging because of the inherent limitations of the soft actuators and artificial muscles that drive them, including low force generation, small actuation range, and poor compactness to name a few. Despite advances in this area, realizing compact soft grippers, which exhibit high dexterity and force output, is still challenging. This article explores using twisted string actuators (TSAs) to drive a soft robotic gripper. TSAs have been widely used in numerous robotic applications, but their inclusion in soft robots has been limited. The proposed design of the gripper was inspired by the human hand, with four fingers and a thumb. Tunable stiffness was implemented in the fingers by using antagonistic TSAs. The fingers' bending angles, actuation speed, blocked force output, and stiffness tuning are experimentally characterized. The gripper achieves a score of 6 on the Kapandji test and recreate 31 of the 33 grasps of the Feix GRASP taxonomy. It exhibits a maximum grasping force of 72 N, which is almost 13 times its own weight. A comparison study reveals that the proposed gripper exhibits equivalent or superior performance compared to other similar soft grippers.