Tingguang Li

Lei Han, Qingxu Zhu, Jiapeng Sheng et al.

Knowledge from animals and humans inspires robotic innovations. Numerous efforts have been made to achieve agile locomotion in quadrupedal robots through classical controllers or reinforcement learning approaches. These methods usually rely on physical models or handcrafted rewards to accurately describe the specific system, rather than on a generalized understanding like animals do. Here we propose a hierarchical framework to construct primitive-, environmental- and strategic-level knowledge that are all pre-trainable, reusable and enrichable for legged robots. The primitive module summarizes knowledge from animal motion data, where, inspired by large pre-trained models in language and image understanding, we introduce deep generative models to produce motor control signals stimulating legged robots to act like real animals. Then, we shape various traversing capabilities at a higher level to align with the environment by reusing the primitive module. Finally, a strategic module is trained focusing on complex downstream tasks by reusing the knowledge from previous levels. We apply the trained hierarchical controllers to the MAX robot, a quadrupedal robot developed in-house, to mimic animals, traverse complex obstacles and play in a designed challenging multi-agent chase tag game, where lifelike agility and strategy emerge in the robots.

Weiqiang Wang, Xuefei Zhe, Qiuhong Ke et al.

We present a neural network-based system for long-term, multi-action human motion synthesis. The system, dubbed as NEURAL MARIONETTE, can produce high-quality and meaningful motions with smooth transitions from simple user input, including a sequence of action tags with expected action duration, and optionally a hand-drawn moving trajectory if the user specifies. The core of our system is a novel Transformer-based motion generation model, namely MARIONET, which can generate diverse motions given action tags. Different from existing motion generation models, MARIONET utilizes contextual information from the past motion clip and future action tag, dedicated to generating actions that can smoothly blend historical and future actions. Specifically, MARIONET first encodes target action tag and contextual information into an action-level latent code. The code is unfolded into frame-level control signals via a time unrolling module, which could be then combined with other frame-level control signals like the target trajectory. Motion frames are then generated in an auto-regressive way. By sequentially applying MARIONET, the system NEURAL MARIONETTE can robustly generate long-term, multi-action motions with the help of two simple schemes, namely "Shadow Start" and "Action Revision". Along with the novel system, we also present a new dataset dedicated to the multi-action motion synthesis task, which contains both action tags and their contextual information. Extensive experiments are conducted to study the action accuracy, naturalism, and transition smoothness of the motions generated by our system.

Haoqing Wang, Xiang Long, Ziheng Li et al.

Reinforcement Learning with Verifiable Rewards (RLVR) plays a key role in stimulating the explicit reasoning capability of Large Language Models (LLMs). We can achieve expert-level performance in some specific domains via RLVR, such as coding or math. When a general multi-domain expert-level model is required, we need to carefully consider the collaboration of RLVR across different domains. The current state-of-the-art models mainly employ two different training paradigms for multi-domain RLVR: mixed multi-task RLVR and separate RLVR followed by model merging. However, most of the works did not provide a detailed comparison and analysis about these paradigms. To this end, we choose multiple commonly used high-level tasks (e.g., math, coding, science, and instruction following) as our target domains and design extensive qualitative and quantitative experiments using open-source datasets. We find the RLVR across domains exhibits few mutual interferences, and reasoning-intensive domains demonstrate mutually synergistic effects. Furthermore, we analyze the internal mechanisms of mutual gains from the perspectives of weight space geometry, model prediction behavior, and information constraints. This project is named as M2RL that means Mixed multi-task training or separate training followed by model Merging for Reinforcement Learning, and the homepage is at https://github.com/mosAI25/M2RL

Sijie Cheng, Kechen Fang, Yangyang Yu et al. · tsinghua

Recent advancements in Multi-modal Large Language Models (MLLMs) have opened new avenues for applications in Embodied AI. Building on previous work, EgoThink, we introduce VidEgoThink, a comprehensive benchmark for evaluating egocentric video understanding capabilities. To bridge the gap between MLLMs and low-level control in Embodied AI, we design four key interrelated tasks: video question-answering, hierarchy planning, visual grounding and reward modeling. To minimize manual annotation costs, we develop an automatic data generation pipeline based on the Ego4D dataset, leveraging the prior knowledge and multimodal capabilities of GPT-4o. Three human annotators then filter the generated data to ensure diversity and quality, resulting in the VidEgoThink benchmark. We conduct extensive experiments with three types of models: API-based MLLMs, open-source image-based MLLMs, and open-source video-based MLLMs. Experimental results indicate that all MLLMs, including GPT-4o, perform poorly across all tasks related to egocentric video understanding. These findings suggest that foundation models still require significant advancements to be effectively applied to first-person scenarios in Embodied AI. In conclusion, VidEgoThink reflects a research trend towards employing MLLMs for egocentric vision, akin to human capabilities, enabling active observation and interaction in the complex real-world environments.

Xiaowei Chi, Kuangzhi Ge, Jiaming Liu et al.

Video Generation Models (VGMs) have become powerful backbones for Vision-Language-Action (VLA) models, leveraging large-scale pretraining for robust dynamics modeling. However, current methods underutilize their distribution modeling capabilities for predicting future states. Two challenges hinder progress: integrating generative processes into feature learning is both technically and conceptually underdeveloped, and naive frame-by-frame video diffusion is computationally inefficient for real-time robotics. To address these, we propose Manipulate in Dream (MinD), a dual-system world model for real-time, risk-aware planning. MinD uses two asynchronous diffusion processes: a low-frequency visual generator (LoDiff) that predicts future scenes and a high-frequency diffusion policy (HiDiff) that outputs actions. Our key insight is that robotic policies do not require fully denoised frames but can rely on low-resolution latents generated in a single denoising step. To connect early predictions to actions, we introduce DiffMatcher, a video-action alignment module with a novel co-training strategy that synchronizes the two diffusion models. MinD achieves a 63% success rate on RL-Bench, 60% on real-world Franka tasks, and operates at 11.3 FPS, demonstrating the efficiency of single-step latent features for control signals. Furthermore, MinD identifies 74% of potential task failures in advance, providing real-time safety signals for monitoring and intervention. This work establishes a new paradigm for efficient and reliable robotic manipulation using generative world models.

Yingtian Tang, Jiangtao Liu, Cheng Zhou et al.

Motion style transfer is highly desired for motion generation systems for gaming. Compared to its offline counterpart, the research on online motion style transfer under interactive control is limited. In this work, we propose an end-to-end neural network that can generate motions with different styles and transfer motion styles in real-time under user control. Our approach eliminates the use of handcrafted phase features, and could be easily trained and directly deployed in game systems. In the experiment part, we evaluate our approach from three aspects that are essential for industrial game design: accuracy, flexibility, and variety, and our model performs a satisfying result.

Jianbang Liu, Baopu Li, Tingguang Li et al.

In this paper, we present a novel path planning algorithm to achieve fast path planning in complex environments. Most existing path planning algorithms are difficult to quickly find a feasible path in complex environments or even fail. However, our proposed framework can overcome this difficulty by using a learning-based prediction module and a sampling-based path planning module. The prediction module utilizes an auto-encoder-decoder-like convolutional neural network (CNN) to output a promising region where the feasible path probably lies in. In this process, the environment is treated as an RGB image to feed in our designed CNN module, and the output is also an RGB image. No extra computation is required so that we can maintain a high processing speed of 60 frames-per-second (FPS). Incorporated with a sampling-based path planner, we can extract a feasible path from the output image so that the robot can track it from start to goal. To demonstrate the advantage of the proposed algorithm, we compare it with conventional path planning algorithms in a series of simulation experiments. The results reveal that the proposed algorithm can achieve much better performance in terms of planning time, success rate, and path length.

Zhaoting Li, Tingguang Li, Jiankun Wang et al.

Being able to explore unknown environments is a requirement for fully autonomous robots. Many learning-based methods have been proposed to learn an exploration strategy. In the frontier-based exploration, learning algorithms tend to learn the optimal or near-optimal frontier to explore. Most of these methods represent the environments as fixed size images and take these as inputs to neural networks. However, the size of environments is usually unknown, which makes these methods fail to generalize to real world scenarios. To address this issue, we present a novel state representation method based on 4D point-clouds-like information, including the locations, frontier, and distance information. We also design a neural network that can process these 4D point-clouds-like information and generate the estimated value for each frontier. Then this neural network is trained using the typical reinforcement learning framework. We test the performance of our proposed method by comparing it with other five methods and test its scalability on a map that is much larger than maps in the training set. The experiment results demonstrate that our proposed method needs shorter average traveling distances to explore whole environments and can be adopted in maps with arbitrarily sizes.

Tingguang Li, Krishnan Srinivasan, Max Qing-Hu Meng et al.

Robotic in-hand manipulation has been a long-standing challenge due to the complexity of modelling hand and object in contact and of coordinating finger motion for complex manipulation sequences. To address these challenges, the majority of prior work has either focused on model-based, low-level controllers or on model-free deep reinforcement learning that each have their own limitations. We propose a hierarchical method that relies on traditional, model-based controllers on the low-level and learned policies on the mid-level. The low-level controllers can robustly execute different manipulation primitives (reposing, sliding, flipping). The mid-level policy orchestrates these primitives. We extensively evaluate our approach in simulation with a 3-fingered hand that controls three degrees of freedom of elongated objects. We show that our approach can move objects between almost all the possible poses in the workspace while keeping them firmly grasped. We also show that our approach is robust to inaccuracies in the object models and to observation noise. Finally, we show how our approach generalizes to objects of other shapes.

Tingguang Li, Weitao Xi, Meng Fang et al.

We present a learning-based approach to solving a Rubik's cube with a multi-fingered dexterous hand. Despite the promising performance of dexterous in-hand manipulation, solving complex tasks which involve multiple steps and diverse internal object structure has remained an important, yet challenging task. In this paper, we tackle this challenge with a hierarchical deep reinforcement learning method, which separates planning and manipulation. A model-based cube solver finds an optimal move sequence for restoring the cube and a model-free cube operator controls all five fingers to execute each move step by step. To train our models, we build a high-fidelity simulator which manipulates a Rubik's Cube, an object containing high-dimensional state space, with a 24-DoF robot hand. Extensive experiments on 1400 randomly scrambled Rubik's cubes demonstrate the effectiveness of our method, achieving an average success rate of 90.3%.

Tingguang Li, Danny Ho, Chenming Li et al.

As one of the most promising areas, mobile robots draw much attention these years. Current work in this field is often evaluated in a few manually designed scenarios, due to the lack of a common experimental platform. Meanwhile, with the recent development of deep learning techniques, some researchers attempt to apply learning-based methods to mobile robot tasks, which requires a substantial amount of data. To satisfy the underlying demand, in this paper we build HouseExpo, a large-scale indoor layout dataset containing 35,126 2D floor plans including 252,550 rooms in total. Together we develop Pseudo-SLAM, a lightweight and efficient simulation platform to accelerate the data generation procedure, thereby speeding up the training process. In our experiments, we build models to tackle obstacle avoidance and autonomous exploration from a learning perspective in simulation as well as real-world experiments to verify the effectiveness of our simulator and dataset. All the data and codes are available online and we hope HouseExpo and Pseudo-SLAM can feed the need for data and benefits the whole community.

Tingguang Li, Jin Pan, Delong Zhu et al.

To achieve scenario intelligence, humans must transfer knowledge to robots by developing goal-oriented algorithms, which are sometimes insensitive to dynamically changing environments. While deep reinforcement learning achieves significant success recently, it is still extremely difficult to be deployed in real robots directly. In this paper, we propose a hybrid structure named Option-Interruption in which human knowledge is embedded into a hierarchical reinforcement learning framework. Our architecture has two key components: options, represented by existing human-designed methods, can significantly speed up the training process and interruption mechanism, based on learnable termination functions, enables our system to quickly respond to the external environment. To implement this architecture, we derive a set of update rules based on policy gradient methods and present a complete training process. In the experiment part, our method is evaluated in Four-room navigation and exploration task, which shows the efficiency and flexibility of our framework.