We propose the expert composer policy, a framework to reliably expand the skill repertoire of quadruped agents. The composer policy links pair of experts via transitions to a sampled target state, allowing experts to be composed sequentially. Each expert specializes in a single skill, such as a locomotion gait or a jumping motion. Instead of a hierarchical or mixture-of-experts architecture, we train a single composer policy in an independent process that is not conditioned on the other expert policies. By reusing the same composer policy, our approach enables adding new experts without affecting existing ones, enabling incremental repertoire expansion and preserving original motion quality. We measured the transition success rate of 72 transition pairs and achieved an average success rate of 99.99%, which is over 10% higher than the baseline random approach, and outperforms other state-of-the-art methods. Using domain randomization during training we ensure a successful transfer to the real world, where we achieve an average transition success rate of 97.22% (N=360) in our experiments.

Reinforcement learning (RL) policies are prone to high-frequency oscillations, especially undesirable when deploying to hardware in the real-world. In this paper, we identify, categorize, and compare methods from the literature that aim to mitigate high-frequency oscillations in deep RL. We define two broad classes: loss regularization and architectural methods. At their core, these methods incentivize learning a smooth mapping, such that nearby states in the input space produce nearby actions in the output space. We present benchmarks in terms of policy performance and control smoothness on traditional RL environments from the Gymnasium and a complex manipulation task, as well as three robotics locomotion tasks that include deployment and evaluation with real-world hardware. Finally, we also propose hybrid methods that combine elements from both loss regularization and architectural methods. We find that the best-performing hybrid outperforms other methods, and improves control smoothness by 26.8% over the baseline, with a worst-case performance degradation of just 2.8%.

This paper proposes the transition-net, a robust transition strategy that expands the versatility of robot locomotion in the real-world setting. To this end, we start by distributing the complexity of different gaits into dedicated locomotion policies applicable to real-world robots. Next, we expand the versatility of the robot by unifying the policies with robust transitions into a single coherent meta-controller by examining the latent state representations. Our approach enables the robot to iteratively expand its skill repertoire and robustly transition between any policy pair in a library. In our framework, adding new skills does not introduce any process that alters the previously learned skills. Moreover, training of a locomotion policy takes less than an hour with a single consumer GPU. Our approach is effective in the real-world and achieves a 19% higher average success rate for the most challenging transition pairs in our experiments compared to existing approaches.

Abstract Pailous are representative Chinese architectural works used for commemoration. However, their geometric structure and semantic construction rules are too complex for quick and intuitive modelling using traditional modelling tools. We propose an intuitive modelling system for the stylized creation of pailous for novices. Our system encapsulates structural components as icons and semantic layouts as topological graphs, using which users create and manipulate icons with topological recommendations. The interpreter automatically and immediately transforms a graph to its corresponding model using built‐in components with the proposed parametric L‐system grammars derived from architectural rules. Using this system to re‐create existing representative pailous and design imaginary ones yields results with the desired visual complexities. In comparison to Maya, a 3D modelling tool, when modelling a pailou and toukung, our system is effective and simple, and eliminates the need to remember and understand complex rules.

The vector graphics with gradient mesh can be attributed to their compactness and scalability; however, they tend to fall short when it comes to real-time editing due to a lack of real-time rasterization and an efficient editing tool for image details. In this paper, we encode global manipulation geometries and local image details within a hybrid vector structure, using parametric patches and detailed features for localized and parallelized thin-plate spline interpolation in order to achieve good compressibility, interactive expressibility, and editability. The proposed system then automatically extracts an optimal set of detailed color features while considering the compression ratio of the image as well as reconstruction error and its characteristics applicable to the preservation of structural and irregular saliency of the image. The proposed real-time vector representation makes it possible to construct an interactive editing system for detail-maintained image magnification and color editing as well as material replacement in cross mapping, without maintaining spatial and temporal consistency while editing in a raster space. Experiments demonstrate that our representation method is superior to several state-of-the-art methods and as good as JPEG, while providing real-time editability and preserving structural and irregular saliency information.

Motion synthesis in a dynamic environment has been a long-standing problem for character animation. Methods using motion capture data tend to scale poorly in complex environments because of their larger capturing and labeling requirement. Physics-based controllers are effective in this regard, albeit less controllable. In this paper, we present CARL, a quadruped agent that can be controlled with high-level directives and react naturally to dynamic environments. Starting with an agent that can imitate individual animation clips, we use Generative Adversarial Networks to adapt high-level controls, such as speed and heading, to action distributions that correspond to the original animations. Further fine-tuning through the deep reinforcement learning enables the agent to recover from unseen external perturbations while producing smooth transitions. It then becomes straightforward to create autonomous agents in dynamic environments by adding navigation modules over the entire process. We evaluate our approach by measuring the agent's ability to follow user control and provide a visual analysis of the generated motion to show its effectiveness.