Embodied agents operating in everyday environments must go beyond explicit instruction following toward proactive assistance, in which they infer latent human needs from ongoing dialogue, ground these needs in the physical scene, and determine whether and how to act. However, existing vision-language-action systems are primarily designed for explicit task execution and struggle to jointly reason over conversational context and dynamic embodied environments. In this work, we introduce HomeSPT-Bench, a benchmark for situated proactive task generation built on the OmniGibson simulator, in which agents must infer implicit needs from multi-party dialogue, ground them in fully specified 3D scenes, and decide on appropriate actions. Evaluations show that state-of-the-art VLM-based agents exhibit a substantial performance gap in this setting, with frequent failures in spatial grounding, occlusion handling, and intervention decision-making. HomeSPT-Bench provides a testbed for studying situated proactive behavior that integrates dialogue understanding with embodied scene grounding, and highlights the remaining gap between language-based reasoning and reliable embodied proactive assistance.

Adversarial attacks and robustness in Deep Reinforcement Learning (DRL) have been widely studied in various threat models; however, few consider environmental state perturbations, which are natural in embodied scenarios. To improve the robustness of DRL agents, we formulate the problem of environmental state perturbation, introducing a preliminary non-targeted attack method as a calibration adversary, and then propose a defense framework, named Boosted Adversarial Training (BAT), which first tunes the agents via supervised learning to avoid catastrophic failure and subsequently adversarially trains the agent with reinforcement learning. Extensive experimental results substantiate the vulnerability of mainstream agents under environmental state perturbations and the effectiveness of our proposed attack. The defense results demonstrate that while existing robust reinforcement learning algorithms may not be suitable, our BAT framework can significantly enhance the robustness of agents against environmental state perturbations across various situations.

Various living creatures exhibit embodiment intelligence, which is reflected by a collaborative interaction of the brain, body, and environment. The actual behavior of embodiment intelligence is generated by a continuous and dynamic interaction between a subject and the environment through information perception and physical manipulation. The physical interaction between a robot and the environment is the basis for realizing embodied perception and learning. Tactile information plays a critical role in this physical interaction process. It can be used to ensure safety, stability, and compliance, and can provide unique information that is difficult to capture using other perception modalities. However, due to the limitations of existing sensors and perception and learning methods, the development of robotic tactile research lags significantly behind other sensing modalities, such as vision and hearing, thereby seriously restricting the development of robotic embodiment intelligence. This paper presents the current challenges related to robotic tactile embodiment intelligence and reviews the theory and methods of robotic embodied tactile intelligence. Tactile perception and learning methods for embodiment intelligence can be designed based on the development of new large-scale tactile array sensing devices, with the aim to make breakthroughs in the neuromorphic computing technology of tactile intelligence.

In this work, we use a deep learning method to tackle the Zero-Shot Learning (ZSL) problem in tactile material recognition by incorporating the advanced semantic information into a training model. Our main technical contribution is our proposal of an end-to-end deep learning framework for solving the tactile ZSL problem. In this framework, we use a Convolutional Neural Network (CNN) to extract the spatial features and Long Short-Term Memory (LSTM) to extract the temporal features in dynamic tactile sequences, and develop a loss function suitable for the ZSL setting. We present the results of experimental evaluations on publicly available datasets, which show the effectiveness of the proposed method.