Minghui Zheng

Sibo Tian, Minghui Zheng, Xiao Liang

Predicting human motion plays a crucial role in ensuring a safe and effective human-robot close collaboration in intelligent remanufacturing systems of the future. Existing works can be categorized into two groups: those focusing on accuracy, predicting a single future motion, and those generating diverse predictions based on observations. The former group fails to address the uncertainty and multi-modal nature of human motion, while the latter group often produces motion sequences that deviate too far from the ground truth or become unrealistic within historical contexts. To tackle these issues, we propose TransFusion, an innovative and practical diffusion-based model for 3D human motion prediction which can generate samples that are more likely to happen while maintaining a certain level of diversity. Our model leverages Transformer as the backbone with long skip connections between shallow and deep layers. Additionally, we employ the discrete cosine transform to model motion sequences in the frequency space, thereby improving performance. In contrast to prior diffusion-based models that utilize extra modules like cross-attention and adaptive layer normalization to condition the prediction on past observed motion, we treat all inputs, including conditions, as tokens to create a more lightweight model compared to existing approaches. Extensive experimental studies are conducted on benchmark datasets to validate the effectiveness of our human motion prediction model.

Minghui Zheng, Hongxu Chen, Huimin Ren et al.

Large language models achieve remarkable performance via extended chain-of-thought (CoT) reasoning, yet this lengthy process incurs substantial inference overhead. Existing CoT compression methods struggle with inflexible manual length budgets, computationally expensive multi-stage training pipelines, and fragile scalability restricted to small models. We propose HMPO (Hybrid Median-length Policy Optimization), a cost-effective, single-stage reinforcement learning framework. HMPO efficiently compresses CoT via three synergistic components: an adaptive median-based budget derived from successful rollouts to eliminate manual tuning, a cosine-decay token reward for smooth length penalization, and a multiplicative reward formulation that substantially mitigates trivial reward hacking by strictly prioritizing answer correctness. Trained exclusively on mathematical data, HMPO generalizes seamlessly across math, code, science, and instruction-following tasks. Extensive experiments scaling from 9B to 122B parameters across dense and Mixture-of-Experts (MoE) architectures demonstrate that HMPO achieves 19%--46% token compression with negligible accuracy degradation, all while drastically reducing training costs compared to existing multi-stage baselines.

Sibo Tian, Minghui Zheng, Xiao Liang

Human motion prediction is a cornerstone of human-robot collaboration (HRC), as robots need to infer the future movements of human workers based on past motion cues to proactively plan their motion, ensuring safety in close collaboration scenarios. The diffusion model has demonstrated remarkable performance in predicting high-quality motion samples with reasonable diversity, but suffers from a slow generative process which necessitates multiple model evaluations, hindering real-world applications. To enable real-time prediction, in this work, we propose training a one-step multi-layer perceptron-based (MLP-based) diffusion model for motion prediction using knowledge distillation and Bayesian optimization. Our method contains two steps. First, we distill a pretrained diffusion-based motion predictor, TransFusion, directly into a one-step diffusion model with the same denoiser architecture. Then, to further reduce the inference time, we remove the computationally expensive components from the original denoiser and use knowledge distillation once again to distill the obtained one-step diffusion model into an even smaller model based solely on MLPs. Bayesian optimization is used to tune the hyperparameters for training the smaller diffusion model. Extensive experimental studies are conducted on benchmark datasets, and our model can significantly improve the inference speed, achieving real-time prediction without noticeable degradation in performance.

Meng-Lun Lee, Sara Behdad, Xiao Liang et al.

Product disassembly is a labor-intensive process and is far from being automated. Typically, disassembly is not robust enough to handle product varieties from different shapes, models, and physical uncertainties due to component imperfections, damage throughout component usage, or insufficient product information. To overcome these difficulties and to automate the disassembly procedure through human-robot collaboration without excessive computational cost, this paper proposes a real-time receding horizon sequence planner that distributes tasks between robot and human operator while taking real-time human motion into consideration. The sequence planner aims to address several issues in the disassembly line, such as varying orientations, safety constraints of human operators, uncertainty of human operation, and the computational cost of large number of disassembly tasks. The proposed disassembly sequence planner identifies both the positions and orientations of the to-be-disassembled items, as well as the locations of human operator, and obtains an optimal disassembly sequence that follows disassembly rules and safety constraints for human operation. Experimental tests have been conducted to validate the proposed planner: the robot can locate and disassemble the components following the optimal sequence, and consider explicitly human operator's real-time motion, and collaborate with the human operator without violating safety constraints.

Zhu Chen, Xiao Liang, Minghui Zheng

Grasping and releasing objects would cause oscillations to delivery drones in the warehouse. To reduce such undesired oscillations, this paper treats the to-be-delivered object as an unknown external disturbance and presents an image-based disturbance observer (DOB) to estimate and reject such disturbance. Different from the existing DOB technique that can only compensate for the disturbance after the oscillations happen, the proposed image-based one incorporates image-based disturbance prediction into the control loop to further improve the performance of the DOB. The proposed image-based DOB consists of two parts. The first one is deep-learning-based disturbance prediction. By taking an image of the to-be-delivered object, a sequential disturbance signal is predicted in advance using a connected pre-trained convolutional neural network (CNN) and a long short-term memory (LSTM) network. The second part is a conventional DOB in the feedback loop with a feedforward correction, which utilizes the deep learning prediction to generate a learning signal. Numerical studies are performed to validate the proposed image-based DOB regarding oscillation reduction for delivery drones during the grasping and releasing periods of the objects.

Wansong Liu, Danyang Luo, Changxu Wu et al.

Currently, studying the vehicle-human interactive behavior in the emergency needs a large amount of datasets in the actual emergent situations that are almost unavailable. Existing public data sources on autonomous vehicles (AVs) mainly focus either on the normal driving scenarios or on emergency situations without human involvement. To fill this gap and facilitate related research, this paper provides a new yet convenient way to extract the interactive behavior data (i.e., the trajectories of vehicles and humans) from actual accident videos that were captured by both the surveillance cameras and driving recorders. The main challenge for data extraction from real-time accident video lies in the fact that the recording cameras are un-calibrated and the angles of surveillance are unknown. The approach proposed in this paper employs image processing to obtain a new perspective which is different from the original video's perspective. Meanwhile, we manually detect and mark object feature points in each image frame. In order to acquire a gradient of reference ratios, a geometric model is implemented in the analysis of reference pixel value, and the feature points are then scaled to the object trajectory based on the gradient of ratios. The generated trajectories not only restore the object movements completely but also reflect changes in vehicle velocity and rotation based on the feature points distributions.