Searcharxiv⌕ Search

arXiv subjects

Takayuki Osa

Publications and source records attributed to Takayuki Osa.

At least 19 recordsLinked to original sources
arXiv2026

Learning Hierarchical Skill Policies with Offline Quality-Diversity Reinforcement Learning

Recent studies investigate how to leverage pre-collected datasets to improve the policy performance and sample efficiency of RL. One promising approach to achieve this goal is to employ a two-stage strategy: In the first stage, diverse skills are extracted as a low-level policy from a given dataset, and a high-level policy is trained to solve a specific task in the second stage. Typically, extraction of the low-level policy is performed based on unsupervised learning such as trajectory VAE. However, a limitation of this approach is that the quality of the low-level policy highly depends on the quality of the dataset. To address this issue, we introduce QDOS (Quality-Diversity Offline Skill learning), a unified pipeline for robust offline-to-online learning. Our approach incorporates an Advantage-Weighted Quality-Diversity pretraining objective, which weights the skill extraction and diversity objectives by the estimated advantage of each trajectory segment. This approach allows the model to extract diverse and high-value skills. By providing robust and task-relevant skill representations, QDOS significantly improves the quality of the embedded skill space used by the low-level policy. We further integrate this with a dual dataset reuse strategy, where offline data is used both for skill pretraining and for populating the online replay buffer via pseudo-labeling. Experiments demonstrate that QDOS significantly outperforms strong baselines in structured manipulation tasks and unstructured locomotion tasks, confirming its ability to accelerate exploration and improve final returns in challenging sparse-reward domains.

cs.AI↗
arXiv2026

Learning the Supports for Categorical Critic in Reinforcement Learning

Value functions are an essential component in actor-critic based deep reinforcement learning (RL). Conventionally, these functions are trained as a regression task by minimising the mean squared error (MSE) relative to bootstrapped target values. Meanwhile, in distributional RL, a distribution of returns is modelled based on the distributional Bellman operator. This work investigates the Gaussian Histogram Loss (HL-Gauss), a recent approach that reframes value estimation as classification by encoding each scalar Bellman target as a Gaussian-smoothed categorical target. Despite its potential, applying histogram-based losses to RL presents inherent challenges, most notably the requirement to pre-define a fixed support interval, which is often complicated by the non-stationary and stochastic nature of target values typically found in RL tasks. In this work, we propose an approach that dynamically learns the lower and upper bounds of the support instead of assigning them beforehand. We derive an objective that jointly learns these bounds whilst learning the categorical representation of the scalar values, and we show that this objective forms an upper bound on the mean-squared Bellman error. Our theoretical analysis further shows that this bound is tighter than that of non-learned supports of HL-Gauss. Empirically, the proposed objective enables stable adaptation of the support interval and matches HL-Gauss-based actor-critic algorithms on most continuous-control tasks whilst improving on a subset, without requiring a pre-specified support interval.

cs.LG↗
arXiv2026

Self-Supervised Theorem Discovery in a Formal Axiomatic System

Recent artificial intelligence (AI) systems have shown remarkable progress in mathematical reasoning. Many existing approaches, including large language models (LLMs), draw on human prior knowledge in the form of mathematical text, code, or theorem libraries. Although these approaches are highly effective in practice, it remains an open question whether an agent can autonomously discover useful theorems without such human priors. We study this question in a formal axiomatic system by developing an agent that starts from axioms and inference rules alone and gradually grows a library of useful theorems. Concretely, we propose a self-supervised theorem-discovery algorithm that alternates between proof search and useful-theorem extraction, building a theorem library whose entries are reused as lemmas for subsequent proof search. Experiments show that the agent discovers tens of thousands of theorems and finds proofs for human-written benchmark problems, suggesting that its discoveries include theorems meaningful from a human mathematical perspective. Furthermore, the discovered theorems improve LLM proof performance when provided as prompt lemmas, indicating that they can serve as external knowledge for LLM reasoning. Our results provide evidence that useful theorems can emerge from proof search without relying on human-provided theorem libraries. More broadly, they suggest a path toward self-evolving AI systems for mathematics whose discoveries remain formally verifiable.

cs.AI↗
arXiv2026

Revisiting Regularized Policy Optimization for Stable and Efficient Reinforcement Learning in Two-Player Games

Two-player games such as board games have long been used as traditional benchmarks for reinforcement learning. This work revisits a policy optimization method with reverse Kullback-Leibler regularization and entropy regularization and analyzes this combination in two-player zero-sum settings from theoretical and empirical perspectives. From a theoretical perspective, we investigate the stability of the policy update rule in two theoretical settings: game-theoretic normal-form games and finite-length games. We provide novel convergence guarantees and verify our theoretical results through numerical experiments on synthetic games. From an empirical perspective, we derive a practical model-free reinforcement learning algorithm based on the regularized policy optimization. We validate the training efficiency of our algorithm through comprehensive experiments on five board games: Animal Shogi, Gardner Chess, Go, Hex, and Othello. Experimental results show that our agent learns more efficiently than existing methods across environments.

cs.LG↗
arXiv2026

Rethinking Policy Diversity in Ensemble Policy Gradient in Large-Scale Reinforcement Learning

Scaling reinforcement learning to tens of thousands of parallel environments requires overcoming the limited exploration capacity of a single policy. Ensemble-based policy gradient methods, which employ multiple policies to collect diverse samples, have recently been proposed to promote exploration. However, merely broadening the exploration space does not always enhance learning capability, since excessive exploration can reduce exploration quality or compromise training stability. In this work, we theoretically analyze the impact of inter-policy diversity on learning efficiency in policy ensembles, and propose Coupled Policy Optimization which regulates diversity through KL constraints between policies. The proposed method enables effective exploration and outperforms strong baselines such as SAPG, PBT, and PPO across multiple tasks, including challenging dexterous manipulation, in terms of both sample efficiency and final performance. Furthermore, analysis of policy diversity and effective sample size during training reveals that follower policies naturally distribute around the leader, demonstrating the emergence of structured and efficient exploratory behavior. Our results indicate that diverse exploration under appropriate regulation is key to achieving stable and sample-efficient learning in ensemble policy gradient methods. Project page at https://naoki04.github.io/paper-cpo/ .

cs.LG↗
arXiv2026

Cross-Embodiment Offline Reinforcement Learning for Heterogeneous Robot Datasets

Scalable robot policy pre-training has been hindered by the high cost of collecting high-quality demonstrations for each platform. In this study, we address this issue by uniting offline reinforcement learning (offline RL) with cross-embodiment learning. Offline RL leverages both expert and abundant suboptimal data, and cross-embodiment learning aggregates heterogeneous robot trajectories across diverse morphologies to acquire universal control priors. We perform a systematic analysis of this offline RL and cross-embodiment paradigm, providing a principled understanding of its strengths and limitations. To evaluate this offline RL and cross-embodiment paradigm, we construct a suite of locomotion datasets spanning 16 distinct robot platforms. Our experiments confirm that this combined approach excels at pre-training with datasets rich in suboptimal trajectories, outperforming pure behavior cloning. However, as the proportion of suboptimal data and the number of robot types increase, we observe that conflicting gradients across morphologies begin to impede learning. To mitigate this, we introduce an embodiment-based grouping strategy in which robots are clustered by morphological similarity and the model is updated with a group gradient. This simple, static grouping substantially reduces inter-robot conflicts and outperforms existing conflict-resolution methods.

cs.AI↗
arXiv2025

Gradual Transition from Bellman Optimality Operator to Bellman Operator in Online Reinforcement Learning

For continuous action spaces, actor-critic methods are widely used in online reinforcement learning (RL). However, unlike RL algorithms for discrete actions, which generally model the optimal value function using the Bellman optimality operator, RL algorithms for continuous actions typically model Q-values for the current policy using the Bellman operator. These algorithms for continuous actions rely exclusively on policy updates for improvement, which often results in low sample efficiency. This study examines the effectiveness of incorporating the Bellman optimality operator into actor-critic frameworks. Experiments in a simple environment show that modeling optimal values accelerates learning but leads to overestimation bias. To address this, we propose an annealing approach that gradually transitions from the Bellman optimality operator to the Bellman operator, thereby accelerating learning while mitigating bias. Our method, combined with TD3 and SAC, significantly outperforms existing approaches across various locomotion and manipulation tasks, demonstrating improved performance and robustness to hyperparameters related to optimality. The code for this study is available at https://github.com/motokiomura/annealed-q-learning.

cs.LG↗
arXiv2025

Offline Reinforcement Learning with Wasserstein Regularization via Optimal Transport Maps

Offline reinforcement learning (RL) aims to learn an optimal policy from a static dataset, making it particularly valuable in scenarios where data collection is costly, such as robotics. A major challenge in offline RL is distributional shift, where the learned policy deviates from the dataset distribution, potentially leading to unreliable out-of-distribution actions. To mitigate this issue, regularization techniques have been employed. While many existing methods utilize density ratio-based measures, such as the $f$-divergence, for regularization, we propose an approach that utilizes the Wasserstein distance, which is robust to out-of-distribution data and captures the similarity between actions. Our method employs input-convex neural networks (ICNNs) to model optimal transport maps, enabling the computation of the Wasserstein distance in a discriminator-free manner, thereby avoiding adversarial training and ensuring stable learning. Our approach demonstrates comparable or superior performance to widely used existing methods on the D4RL benchmark dataset. The code is available at https://github.com/motokiomura/Q-DOT .

cs.LG↗
arXiv2024

Stabilizing Extreme Q-learning by Maclaurin Expansion

In offline reinforcement learning, in-sample learning methods have been widely used to prevent performance degradation caused by evaluating out-of-distribution actions from the dataset. Extreme Q-learning (XQL) employs a loss function based on the assumption that Bellman error follows a Gumbel distribution, enabling it to model the soft optimal value function in an in-sample manner. It has demonstrated strong performance in both offline and online reinforcement learning settings. However, issues remain, such as the instability caused by the exponential term in the loss function and the risk of the error distribution deviating from the Gumbel distribution. Therefore, we propose Maclaurin Expanded Extreme Q-learning to enhance stability. In this method, applying Maclaurin expansion to the loss function in XQL enhances stability against large errors. This approach involves adjusting the modeled value function between the value function under the behavior policy and the soft optimal value function, thus achieving a trade-off between stability and optimality depending on the order of expansion. It also enables adjustment of the error distribution assumption from a normal distribution to a Gumbel distribution. Our method significantly stabilizes learning in online RL tasks from DM Control, where XQL was previously unstable. Additionally, it improves performance in several offline RL tasks from D4RL.

cs.LG↗
arXiv2024

Discovering Multiple Solutions from a Single Task in Offline Reinforcement Learning

Recent studies on online reinforcement learning (RL) have demonstrated the advantages of learning multiple behaviors from a single task, as in the case of few-shot adaptation to a new environment. Although this approach is expected to yield similar benefits in offline RL, appropriate methods for learning multiple solutions have not been fully investigated in previous studies. In this study, we therefore addressed the problem of finding multiple solutions from a single task in offline RL. We propose algorithms that can learn multiple solutions in offline RL, and empirically investigate their performance. Our experimental results show that the proposed algorithm learns multiple qualitatively and quantitatively distinctive solutions in offline RL.

cs.LG↗
arXiv2024

Offline Reinforcement Learning from Datasets with Structured Non-Stationarity

Current Reinforcement Learning (RL) is often limited by the large amount of data needed to learn a successful policy. Offline RL aims to solve this issue by using transitions collected by a different behavior policy. We address a novel Offline RL problem setting in which, while collecting the dataset, the transition and reward functions gradually change between episodes but stay constant within each episode. We propose a method based on Contrastive Predictive Coding that identifies this non-stationarity in the offline dataset, accounts for it when training a policy, and predicts it during evaluation. We analyze our proposed method and show that it performs well in simple continuous control tasks and challenging, high-dimensional locomotion tasks. We show that our method often achieves the oracle performance and performs better than baselines.

cs.LG↗
arXiv2024

Robustifying a Policy in Multi-Agent RL with Diverse Cooperative Behaviors and Adversarial Style Sampling for Assistive Tasks

Autonomous assistance of people with motor impairments is one of the most promising applications of autonomous robotic systems. Recent studies have reported encouraging results using deep reinforcement learning (RL) in the healthcare domain. Previous studies showed that assistive tasks can be formulated as multi-agent RL, wherein there are two agents: a caregiver and a care-receiver. However, policies trained in multi-agent RL are often sensitive to the policies of other agents. In such a case, a trained caregiver's policy may not work for different care-receivers. To alleviate this issue, we propose a framework that learns a robust caregiver's policy by training it for diverse care-receiver responses. In our framework, diverse care-receiver responses are autonomously learned through trials and errors. In addition, to robustify the care-giver's policy, we propose a strategy for sampling a care-receiver's response in an adversarial manner during the training. We evaluated the proposed method using tasks in an Assistive Gym. We demonstrate that policies trained with a popular deep RL method are vulnerable to changes in policies of other agents and that the proposed framework improves the robustness against such changes.

cs.RO↗
arXiv2024

Symmetric Q-learning: Reducing Skewness of Bellman Error in Online Reinforcement Learning

In deep reinforcement learning, estimating the value function to evaluate the quality of states and actions is essential. The value function is often trained using the least squares method, which implicitly assumes a Gaussian error distribution. However, a recent study suggested that the error distribution for training the value function is often skewed because of the properties of the Bellman operator, and violates the implicit assumption of normal error distribution in the least squares method. To address this, we proposed a method called Symmetric Q-learning, in which the synthetic noise generated from a zero-mean distribution is added to the target values to generate a Gaussian error distribution. We evaluated the proposed method on continuous control benchmark tasks in MuJoCo. It improved the sample efficiency of a state-of-the-art reinforcement learning method by reducing the skewness of the error distribution.

cs.LG↗
arXiv2022

Discovering Diverse Solutions in Deep Reinforcement Learning by Maximizing State-Action-Based Mutual Information

Reinforcement learning algorithms are typically limited to learning a single solution for a specified task, even though diverse solutions often exist. Recent studies showed that learning a set of diverse solutions is beneficial because diversity enables robust few-shot adaptation. Although existing methods learn diverse solutions by using the mutual information as unsupervised rewards, such an approach often suffers from the bias of the gradient estimator induced by value function approximation. In this study, we propose a novel method that can learn diverse solutions without suffering the bias problem. In our method, a policy conditioned on a continuous or discrete latent variable is trained by directly maximizing the variational lower bound of the mutual information, instead of using the mutual information as unsupervised rewards as in previous studies. Through extensive experiments on robot locomotion tasks, we demonstrate that the proposed method successfully learns an infinite set of diverse solutions by learning continuous latent variables, which is more challenging than learning a finite number of solutions. Subsequently, we show that our method enables more effective few-shot adaptation compared with existing methods.

stat.ML↗
arXiv2021

Meta-Model-Based Meta-Policy Optimization

Model-based meta-reinforcement learning (RL) methods have recently been shown to be a promising approach to improving the sample efficiency of RL in multi-task settings. However, the theoretical understanding of those methods is yet to be established, and there is currently no theoretical guarantee of their performance in a real-world environment. In this paper, we analyze the performance guarantee of model-based meta-RL methods by extending the theorems proposed by Janner et al. (2019). On the basis of our theoretical results, we propose Meta-Model-Based Meta-Policy Optimization (M3PO), a model-based meta-RL method with a performance guarantee. We demonstrate that M3PO outperforms existing meta-RL methods in continuous-control benchmarks.

cs.LG↗
arXiv2021

Motion Planning by Learning the Solution Manifold in Trajectory Optimization

The objective function used in trajectory optimization is often non-convex and can have an infinite set of local optima. In such cases, there are diverse solutions to perform a given task. Although there are a few methods to find multiple solutions for motion planning, they are limited to generating a finite set of solutions. To address this issue, we presents an optimization method that learns an infinite set of solutions in trajectory optimization. In our framework, diverse solutions are obtained by learning latent representations of solutions. Our approach can be interpreted as training a deep generative model of collision-free trajectories for motion planning. The experimental results indicate that the trained model represents an infinite set of homotopic solutions for motion planning problems.

cs.RO↗
arXiv2020

Goal-Conditioned Variational Autoencoder Trajectory Primitives with Continuous and Discrete Latent Codes

Imitation learning is an intuitive approach for teaching motion to robotic systems. Although previous studies have proposed various methods to model demonstrated movement primitives, one of the limitations of existing methods is that the shape of the trajectories are encoded in high dimensional space. The high dimensionality of the trajectory representation can be a bottleneck in the subsequent process such as planning a sequence of primitive motions. We address this problem by learning the latent space of the robot trajectory. If the latent variable of the trajectories can be learned, it can be used to tune the trajectory in an intuitive manner even when the user is not an expert. We propose a framework for modeling demonstrated trajectories with a neural network that learns the low-dimensional latent space. Our neural network structure is built on the variational autoencoder (VAE) with discrete and continuous latent variables. We extend the structure of the existing VAE to obtain the decoder that is conditioned on the goal position of the trajectory for generalization to different goal positions. Although the inference performed by VAE is not accurate, the positioning error at the generalized goal position can be reduced to less than 1~mm by incorporating the projection onto the solution space. To cope with requirement of the massive training data, we use a trajectory augmentation technique inspired by the data augmentation commonly used in the computer vision community. In the proposed framework, the latent variables that encodes the multiple types of trajectories are learned in an unsupervised manner, although existing methods usually require label information to model diverse behaviors. The learned decoder can be used as a motion planner in which the user can specify the goal position and the trajectory types by setting the latent variables.

cs.RO↗
arXiv2020

Multimodal Trajectory Optimization for Motion Planning

Existing motion planning methods often have two drawbacks: 1) goal configurations need to be specified by a user, and 2) only a single solution is generated under a given condition. In practice, multiple possible goal configurations exist to achieve a task. Although the choice of the goal configuration significantly affects the quality of the resulting trajectory, it is not trivial for a user to specify the optimal goal configuration. In addition, the objective function used in the trajectory optimization is often non-convex, and it can have multiple solutions that achieve comparable costs. In this study, we propose a framework that determines multiple trajectories that correspond to the different modes of the cost function. We reduce the problem of identifying the modes of the cost function to that of estimating the density induced by a distribution based on the cost function. The proposed framework enables users to select a preferable solution from multiple candidate trajectories, thereby making it easier to tune the cost function and obtain a satisfactory solution. We evaluated our proposed method with motion planning tasks in 2D and 3D space. Our experiments show that the proposed algorithm is capable of determining multiple solutions for those tasks.

cs.RO↗