Abstract--Vision-based Proximal Policy Optimization (PPO) struggles with visual observation-based robotic laparoscopic surgical tasks due to the high-dimensional nature of visual input, the sparsity of rewards in surgical environments, and the difficulty of extracting task-relevant features from raw visual data. We introduce a simple approach integrating MedFlamingo, a medical domain-specific Vision-Language Model, with PPO. Our method is evaluated on five diverse laparoscopic surgery task environments in LapGym, using only endoscopic visual observations. MedFlamingo PPO outperforms and converges faster compared to both standard vision-based PPO and OpenFlamingo PPO baselines, achieving task success rates exceeding 70% across all environments, with improvements ranging from 66.67% to 1114.29% compared to baseline. By processing task observations and instructions once per episode to generate high-level planning tokens, our method efficiently combines medical expertise with real-time visual feedback. Our results highlight the value of specialized medical knowledge in robotic surgical planning and decision-making.

In the field of robotic manipulation, the proficiency of deformable object manipulation lags behind human capabilities due to the inherent characteristics of deformable objects. These objects have infinite degrees of freedom, resulting in non-trivial perception and state estimation, and complex dynamics, complicating the prediction of future configurations. Although recent research has focused on deformable object manipulation, most approaches rely on static vision and simple manipulation techniques, limiting the performance level. This paper proposes two solutions to enhance the performance: interactive perception and the use of assistive tools. The first solution posits that optimal perspectives exist during deformable object manipulation, facilitating easier state estimation. By exploring the action-perception regularity, interactive perception facilitates better manipulation and perception. The second solution advocates for the use of assistive tools, a hallmark of human intelligence, to improve manipulation performance. For instance, a folding board can aid in garment folding tasks by reducing object deformation and managing complex dynamics. Hence, this research aims to address the deformable object manipulation problem by incorporating interactive perception and assistive tools to augment manipulation performance.

Deformable object manipulation (DOM) is a long-standing challenge in robotics and has attracted significant interest recently. This paper presents DaXBench, a differentiable simulation framework for DOM. While existing work often focuses on a specific type of deformable objects, DaXBench supports fluid, rope, cloth...; it provides a general-purpose benchmark to evaluate widely different DOM methods, including planning, imitation learning, and reinforcement learning. DaXBench combines recent advances in deformable object simulation with JAX, a high-performance computational framework. All DOM tasks in DaXBench are wrapped with the OpenAI Gym API for easy integration with DOM algorithms. We hope that DaXBench provides to the research community a comprehensive, standardized benchmark and a valuable tool to support the development and evaluation of new DOM methods. Deformable object manipulation (DOM) is a crucial area of research with broad applications, from household (Maitin-Shepard et al., 2010; Miller et al., 2011; Ma et al., 2022) to industrial settings (Miller et al., 2012; Zhu et al., 2022). To aid in algorithm development and prototyping, several DOM benchmarks (Lin et al., 2021; Huang et al., 2021) have been developed using deformable object simulators. However, the high dimensional state and action spaces remain a significant challenge to DOM. Differentiable physics is a promising direction for developing control policies for deformable objects. It implements physical laws as differentiable computational graphs (Freeman et al., 2021; Hu et al., 2020), enabling the optimization of control policies with analytical gradients and therefore improving sample efficiency. Recent studies have shown that differentiable physics-based DOM methods can benefit greatly from this approach (Huang et al., 2021; Heiden et al., 2021; Xu et al., 2022; Chen et al., 2023).

In this paper, we propose a novel approach to 3D deformable object manipulation leveraging a deep neural network called DeformerNet. Controlling the shape of a 3D object requires an effective state representation that can capture the full 3D geometry of the object. Current methods work around this problem by defining a set of feature points on the object or only deforming the object in 2D image space, which does not truly address the 3D shape control problem. Instead, we explicitly use 3D point clouds as the state representation and apply Convolutional Neural Network on point clouds to learn the 3D features. These features are then mapped to the robot end-effector's position using a fully-connected neural network. Once trained in an end-to-end fashion, DeformerNet directly maps the current point cloud of a deformable object, as well as a target point cloud shape, to the desired displacement in robot gripper position. In addition, we investigate the problem of predicting the manipulation point location given the initial and goal shape of the object.

We present a novel approach to deformable object manipulation that does not rely on highly-accurate modeling. The key contribution of this paper is to formulate the task as a Multi-Armed Bandit problem, with each arm representing a model of the deformable object. To "pull" an arm and evaluate its utility, we use the arm's model to generate a velocity command for the gripper(s) holding the object and execute it. As the task proceeds and the object deforms, the utility of each model can change. Our framework estimates these changes and balances exploration of the model set with exploitation of high-utility models. We also propose an approach based on Kalman Filtering for Non-stationary Multi-armed Normal Bandits (KF-MANB) to leverage the coupling between models to learn more from each arm pull. We demonstrate that our method outperforms previous methods on synthetic trials, and performs competitively on several manipulation tasks in simulation.