Advances in robot learning have enabled robots to generate skills for a variety of tasks. Yet, robot learning is typically sample inefficient, struggles to learn from data sources exhibiting varied behaviors, and does not naturally incorporate constraints. These properties are critical for fast, agile tasks such as playing table tennis. Modern techniques for learning from demonstration improve sample efficiency and scale to diverse data, but are rarely evaluated on agile tasks. In the case of reinforcement learning, achieving good performance requires training on high-fidelity simulators. To overcome these limitations, we develop a novel diffusion modeling approach that is offline, constraint-guided, and expressive of diverse agile behaviors. The key to our approach is a kinematic constraint gradient guidance (KCGG) technique that computes gradients through both the forward kinematics of the robot arm and the diffusion model to direct the sampling process. KCGG minimizes the cost of violating constraints while simultaneously keeping the sampled trajectory in-distribution of the training data. We demonstrate the effectiveness of our approach for time-critical robotic tasks by evaluating KCGG in two challenging domains: simulated air hockey and real table tennis. In simulated air hockey, we achieved a 25.4% increase in block rate, while in table tennis, we achieved a 17.3% increase in success rate compared to imitation learning baselines.

In this paper, we propose a novel and generalizable zero-shot knowledge transfer framework that distills expert sports navigation strategies from web videos into robotic systems with adversarial constraints and out-of-distribution image trajectories. Our pipeline enables diffusion-based imitation learning by reconstructing the full 3D task space from multiple partial views, warping it into 2D image space, closing the planning loop within this 2D space, and transfer constrained motion of interest back to task space. Additionally, we demonstrate that the learned policy can serve as a local planner in conjunction with position control. We apply this framework in the wheelchair tennis navigation problem to guide the wheelchair into the ball-hitting region. Our pipeline achieves a navigation success rate of$\mathbf{9 7. 6 7 \%}$in reaching real-world recorded tennis ball trajectories with a physical robot wheelchair, and achieve a success rate of 68.49% in a real-world, real-time experiment on a full-sized tennis court22Code is at https://github.gatech.edu/MCG-Lab/tennis_gameplay_learning.

Robots in dynamic environments need fast, accurate models of how objects move in their environments to support agile planning. In sports such as ping pong, analytical models often struggle to accurately predict ball trajectories with spins due to complex aerodynamics, elastic behaviors, and the challenges of modeling sliding and rolling friction. On the other hand, despite the promise of data-driven methods, machine learning struggles to make accurate, consistent predictions without precise input. In this paper, we propose an end-to-end learning framework that can jointly train a dynamics model and a factor graph estimator. Our approach leverages a Gram-Schmidt (GS) process to extract roto-translational invariant representations to improve the model performance, which can further reduce the validation error compared to data augmentation method. Additionally, we propose a network architecture that enhances nonlinearity by using self-multiplicative bypasses in the layer connections. By leveraging these novel methods, our proposed approach predicts the ball's position with an RMSE of 37.2 mm at the apex after the first bounce, and 71.5 mm after the second bounce.

Advancements in high-throughput sequencing technologies and artificial intelligence (AI) offer unprecedented opportunities for groundbreaking discoveries in bioinformatics research. However, the challenges of exponential growth of omics data and the rapid development of AI technologies require automated big biological data analysis capability and interdisciplinary knowledge-driven scientific insight. Here, we propose a data-intelligence-intensive bioinformatics copilot (Bio-Copilot) system that synergizes AI capabilities with human researchers to facilitate hypothesis-free exploratory research and inspire novel scientific insights in large-scale omics studies. Bio-Copilot forms high-quality intensive intelligence through close collaboration between multiple agents, driven by large language models (LLMs), and human researchers. To augment the capabilities of Bio-Copilot, this study devises an agent group management strategy, an effective human-agent interaction mechanism, a shared interdisciplinary knowledge database, and continuous learning strategies for the agents. We comprehensively compare Bio-Copilot against GPT-4o and several leading AI agents across diverse bioinformatics tasks, using a broad range of evaluation metrics. Bio-Copilot achieves overall state-of-the-art performance across all tasks, while showcasing exceptional task completeness. Furthermore, on application to constructing a large-scale human lung cell atlas, Bio-Copilot not only reproduces the intricate data integration process detailed in a seminal study but also introduces a recursive, multilevel annotation strategy to capture the continuous nature of cellular states and uncovers the characteristics of rare cell types, highlighting its potential to unravel hidden complexities in biological systems. Beyond the technical achievements, this study also underscores the profound implications of integrating AI capabilities with expert knowledge in accelerating impactful biological discoveries and exploring uncharted territories.

The rapid and precise localization and prediction of a ball are critical for developing agile robots in ball sports, particularly in sports like tennis characterized by high-speed ball movements and powerful spins. The Magnus effect induced by spin adds complexity to trajectory prediction during flight and bounce dynamics upon contact with the ground. In this study, we introduce an innovative approach that combines a multi-camera system with factor graphs for real-time and asynchronous 3D tennis ball localization. Additionally, we estimate hidden states like velocity and spin for trajectory prediction. Furthermore, to enhance spin inference early in the ball’s flight, where limited observations are available, we integrate human pose data using a temporal convolutional network (TCN) to compute spin priors within the factor graph. This refinement provides more accurate spin priors at the beginning of the factor graph, leading to improved early-stage hidden state inference for prediction. Our results show the trained TCN can predict the spin priors with RMSE of 5.27 Hz. Integrating TCN into the factor graph reduces the prediction error of landing positions by over 63.6% compared to a baseline method that utilized an adaptive extended Kalman filter.

Client-side adaptive bitrate (ABR) algorithms based on deep reinforcement learning (RL) can continuously improve its adaptability to network conditions. However, most existing methods adopt fixed reward functions to train the ABR policy, which leads the results being not consistent with user-perceived quality of experience (QoE) in a long duration under various network conditions. In order to optimize the QoE, this paper proposes a novel ABR algorithm considering user preference based on short trajectory segments. The user-specific preference feedback, which is selected by the user from a pair of short track segments in advance, is collected and applied to define the training goal of RL. Specifically, we train a deep neural network to define the RL reward and integrate it with A3C-based ABR algorithm. The experiment results show that the accuracy of the proposed reward model outperforms most existing fixed reward functions by 13.6% in user preference prediction, and the optimized ABR algorithm improves QoE by 16.4% on average.