By pairing historical and future trajectories and applying contrastive learning on the encoded feature space, we enforce same-space consistency constraints.

We consider offline reinforcement learning (RL) with heterogeneous agents under severe data scarcity, i.e., we only observe a single historical trajectory for every agent under an unknown, potentially sub-optimal policy. We find that the performance of state-of-the-art offline and model-based RL methods degrade significantly given such limited data availability, even for commonly perceived solved benchmark settings such as MountainCar and CartPole. To address this challenge, we propose PerSim, a model-based offline RL approach which first learns a personalized simulator for each agent by collectively using the historical trajectories across all agents, prior to learning a policy. We do so by positing that the transition dynamics across agents can be represented as a latent function of latent factors associated with agents, states, and actions; subsequently, we theoretically establish that this function is well-approximated by a low-rank decomposition of separable agent, state, and action latent functions. This representation suggests a simple, regularized neural network architecture to effectively learn the transition dynamics per agent, even with scarce, offline data. We perform extensive experiments across several benchmark environments and RL methods. The consistent improvement of our approach, measured in terms of both state dynamics prediction and eventual reward, confirms the efficacy of our framework in leveraging limited historical data to simultaneously learn personalized policies across agents.

By pairing historical and future trajectories and applying contrastive learning on the encoded feature space, we enforce same-space consistency constraints.

The popular path query - identifying the most frequented routes between locations from historical trajectory data - has important applications in urban planning, navigation optimization, and travel recommendations. While traditional algorithms and machine learning approaches have achieved success in this domain, they typically require model training, parameter tuning, and retraining when accommodating data updates. As Large Language Models (LLMs) demonstrate increasing capabilities in spatial and graph-based reasoning, there is growing interest in exploring how these models can be applied to geo-spatial problems. We introduce CompassLLM, a novel multi-agent framework that intelligently leverages the reasoning capabilities of LLMs into the geo-spatial domain to solve the popular path query. CompassLLM employs its agents in a two-stage pipeline: the SEARCH stage that identifies popular paths, and a GENERATE stage that synthesizes novel paths in the absence of an existing one in the historical trajectory data. Experiments on real and synthetic datasets show that CompassLLM demonstrates superior accuracy in SEARCH and competitive performance in GENERATE while being cost-effective.

Next location prediction plays a critical role in understanding human mobility patterns. However, existing approaches face two core limitations: (1) they fall short in capturing the complex, multi-functional semantics of real-world locations; and (2) they lack the capacity to model heterogeneous behavioral dynamics across diverse user groups. To tackle these challenges, we introduce NextLocMoE, a novel framework built upon large language models (LLMs) and structured around a dual-level Mixture-of-Experts (MoE) design. Our architecture comprises two specialized modules: a Location Semantics MoE that operates at the embedding level to encode rich functional semantics of locations, and a Personalized MoE embedded within the Transformer backbone to dynamically adapt to individual user mobility patterns. In addition, we incorporate a history-aware routing mechanism that leverages long-term trajectory data to enhance expert selection and ensure prediction stability. Empirical evaluations across several real-world urban datasets show that NextLocMoE achieves superior performance in terms of predictive accuracy, cross-domain generalization, and interpretability

While reinforcement learning (RL) has achieved notable success in various domains, training effective policies for complex tasks remains challenging. Agents often converge to local optima and fail to maximize long-term rewards. Existing approaches to mitigate training bottlenecks typically fall into two categories: (i) Automated policy refinement, which identifies critical states from past trajectories to guide policy updates, but suffers from costly and uncertain model training; and (ii) Human-in-the-loop refinement, where human feedback is used to correct agent behavior, but this does not scale well to environments with large or continuous action spaces. In this work, we design a large language model-guided policy modulation framework that leverages LLMs to improve RL training without additional model training or human intervention. We first prompt an LLM to identify critical states from a sub-optimal agent's trajectories. Based on these states, the LLM then provides action suggestions and assigns implicit rewards to guide policy refinement. Experiments across standard RL benchmarks demonstrate that our method outperforms state-of-the-art baselines, highlighting the effectiveness of LLM-based explanations in addressing RL training bottlenecks.

Nonprehensile manipulation is crucial for handling objects that are too thin, large, or otherwise ungraspable in unstructured environments. While conventional planning-based approaches struggle with complex contact modeling, learning-based methods have recently emerged as a promising alternative. However, existing learning-based approaches face two major limitations: they heavily rely on multi-view cameras and precise pose tracking, and they fail to generalize across varying physical conditions, such as changes in object mass and table friction. To address these challenges, we propose the Dynamics-Adaptive World Action Model (DyWA), a novel framework that enhances action learning by jointly predicting future states while adapting to dynamics variations based on historical trajectories. By unifying the modeling of geometry, state, physics, and robot actions, DyWA enables more robust policy learning under partial observability. Compared to baselines, our method improves the success rate by 31.5% using only single-view point cloud observations in the simulation. Furthermore, DyWA achieves an average success rate of 68% in real-world experiments, demonstrating its ability to generalize across diverse object geometries, adapt to varying table friction, and robustness in challenging scenarios such as half-filled water bottles and slippery surfaces.

Pedestrian trajectory prediction is important in the research of mobile robot navigation in environments with pedestrians. Most pedestrian trajectory prediction algorithms require the input historical trajectories to be complete. If a pedestrian is unobservable in any frame in the past, then its historical trajectory become incomplete, the algorithm will not predict its future trajectory. To address this limitation, we propose the STGN-IT, a spatio-temporal graph network allowing incomplete trajectory input, which can predict the future trajectories of pedestrians with incomplete historical trajectories. STGN-IT uses the spatio-temporal graph with an additional encoding method to represent the historical trajectories and observation states of pedestrians. Moreover, STGN-IT introduces static obstacles in the environment that may affect the future trajectories as nodes to further improve the prediction accuracy. A clustering algorithm is also applied in the construction of spatio-temporal graphs. Experiments on public datasets show that STGN-IT outperforms state of the art algorithms on these metrics.

We consider offline reinforcement learning (RL) with heterogeneous agents under severe data scarcity, i.e., we only observe a single historical trajectory for every agent under an unknown, potentially sub-optimal policy. We find that the performance of state-of-the-art offline and model-based RL methods degrade significantly given such limited data availability, even for commonly perceived "solved" benchmark settings such as "MountainCar" and "CartPole". To address this challenge, we propose PerSim, a model-based offline RL approach which first learns a personalized simulator for each agent by collectively using the historical trajectories across all agents, prior to learning a policy. We do so by positing that the transition dynamics across agents can be represented as a latent function of latent factors associated with agents, states, and actions; subsequently, we theoretically establish that this function is well-approximated by a "low-rank" decomposition of separable agent, state, and action latent functions. This representation suggests a simple, regularized neural network architecture to effectively learn the transition dynamics per agent, even with scarce, offline data.

Human mobility studies how people move to access their needed resources and plays a significant role in urban planning and location-based services. As a paramount task of human mobility modeling, next location prediction is challenging because of the diversity of users' historical trajectories that gives rise to complex mobility patterns and various contexts. Deep sequential models have been widely used to predict the next location by leveraging the inherent sequentiality of trajectory data. However, they do not fully leverage the relationship between locations and fail to capture users' multi-level preferences. This work constructs a trajectory graph from users' historical traces and proposes a \textbf{Traj}ectory \textbf{G}raph \textbf{E}nhanced \textbf{O}rientation-based \textbf{S}equential network (TrajGEOS) for next-location prediction tasks. TrajGEOS introduces hierarchical graph convolution to capture location and user embeddings. Such embeddings consider not only the contextual feature of locations but also the relation between them, and serve as additional features in downstream modules. In addition, we design an orientation-based module to learn users' mid-term preferences from sequential modeling modules and their recent trajectories. Extensive experiments on three real-world LBSN datasets corroborate the value of graph and orientation-based modules and demonstrate that TrajGEOS outperforms the state-of-the-art methods on the next location prediction task.