Summary The paper proposes a differentiable pipeline that can jointly learn a latent space representation (via variational autoencoder) for controlling soft robots and optimize for the controller and the soft robot parameters for tasks in simulation, such as making a soft 2D robot walk forward as fast as possible. The work is made possible by using a differentiable hybrid-particle-grid based soft material physics simulator. The authors provided insightful details on the alternative minimization scheme for training the autoencoder, the controller neural network, and the robot parameters in tandem. The proposed framework was evaluated on 5 simulated experiments that show controllers using the learned representation outperforms ones using the baseline representation obtained via k-means clustering. Review While the performance of the system is impressive, the motivation of the approach is not well-communicated in 3 folds: In discussing the proposed hybrid-particle-grid based soft robot representation vs finite element methods, the authors claim that the high "degrees of freedom of finite element methods is impractical for most modern control algorithms."

In visual-based Reinforcement Learning (RL), agents often struggle to generalize well to environmental variations in the state space that were not observed during training. The variations can arise in both task-irrelevant features, such as background noise, and task-relevant features, such as robot configurations, that are related to the optimal decisions. To achieve generalization in both situations, agents are required to accurately understand the impact of changed features on the decisions, i.e., establishing the true associations between changed features and decisions in the policy model. However, due to the inherent correlations among features in the state space, the associations between features and decisions become entangled, making it difficult for the policy to distinguish them. To this end, we propose Saliency-Guided Features Decorrelation (SGFD) to eliminate these correlations through sample reweighting. Concretely, SGFD consists of two core techniques: Random Fourier Functions (RFF) and the saliency map. RFF is utilized to estimate the complex non-linear correlations in high-dimensional images, while the saliency map is designed to identify the changed features. Under the guidance of the saliency map, SGFD employs sample reweighting to minimize the estimated correlations related to changed features, thereby achieving decorrelation in visual RL tasks. Our experimental results demonstrate that SGFD can generalize well on a wide range of test environments and significantly outperforms state-of-the-art methods in handling both task-irrelevant variations and task-relevant variations.

However, the limited field-of-view (FoV) of the microscopic camera necessitates camera motion to capture a broader workspace environment. In this work, we propose an autonomous robotic control method to constrain a robot-held camera within a designated FoV. Furthermore, we model the camera extrinsics as part of the kinematic model and use camera measurements coupled with a U-Net based tool tracking to adapt the complete robotic model during task execution. As a proof-of-concept demonstration, the proposed framework was evaluated in a bi-manual setup, where the microscopic camera was controlled to view a tool moving in a pre-defined trajectory. The proposed method allowed the camera to stay 99.5% of the time within the real FoV, compared to 48.1% without the proposed adaptive control.

Soft, growing inflated beam robots, also known as everting vine robots, have previously been shown to navigate confined spaces with ease. Less is known about their ability to navigate three-dimensional open spaces where they have the potential to collapse under their own weight as they attempt to move through a space. Previous work has studied collapse of inflated beams and vine robots due to purely transverse or purely axial external loads. Here, we extend previous models to predict the length at which straight vine robots will collapse under their own weight at arbitrary launch angle relative to gravity, inflated diameter, and internal pressure. Our model successfully predicts the general trends of collapse behavior of straight vine robots. We find that collapse length increases non-linearly with the robot's launch angle magnitude, linearly with the robot's diameter, and with the square root of the robot's internal pressure. We also demonstrate the use of our model to determine the robot parameters required to grow a vine robot across a gap in the floor. This work forms the foundation of an approach for modeling the collapse of vine robots and inflated beams in arbitrary shapes.

Machine learning techniques rely on large and diverse datasets for generalization. Computer vision, natural language processing, and other applications can often reuse public datasets to train many different models. However, due to differences in physical configurations, it is challenging to leverage public datasets for training robotic control policies on new robot platforms or for new tasks. In this work, we propose a novel framework, ExAug to augment the experiences of different robot platforms from multiple datasets in diverse environments. ExAug leverages a simple principle: by extracting 3D information in the form of a point cloud, we can create much more complex and structured augmentations, utilizing both generating synthetic images and geometric-aware penalization that would have been suitable in the same situation for a different robot, with different size, turning radius, and camera placement. The trained policy is evaluated on two new robot platforms with three different cameras in indoor and outdoor environments with obstacles.