Neural Information Processing Systems http://nips.cc/

This paper studies a number of variations on the topic of training a deep network using data generated by a Monte-Carlo Tree Search (MCTS) agent. The paper focuses on the Atari 2600 platform and is motivated by the observation that, while MCTS performs extremely well on Atari 2600 games, it is also too computationally expensive to be used in a realistic setting. The authors provide empirical results on a number of Atari 2600 games.

Neural Information Processing Systems http://nips.cc/

This paper examines replication portfolio construction in incomplete markets - a key problem in financial engineering with applications in pricing, hedging, balance sheet management, and energy storage planning. We model this as a two-player game between an investor and the market, where the investor makes strategic bets on future states while the market reveals outcomes. Inspired by the success of Monte Carlo Tree Search in stochastic games, we introduce an AlphaZero-based system and compare its performance to deep hedging - a widely used industry method based on gradient descent. Through theoretical analysis and experiments, we show that deep hedging struggles in environments where the $Q$-function is not subject to convexity constraints - such as those involving non-convex transaction costs, capital constraints, or regulatory limitations - converging to local optima. We construct specific market environments to highlight these limitations and demonstrate that AlphaZero consistently finds near-optimal replication strategies. On the theoretical side, we establish a connection between deep hedging and convex optimization, suggesting that its effectiveness is contingent on convexity assumptions. Our experiments further suggest that AlphaZero is more sample-efficient - an important advantage in data-scarce, overfitting-prone derivative markets.

Multi-step manipulation in dynamic environments remains challenging. Two major families of methods fail in distinct ways: (i) imitation learning (IL) is reactive but lacks compositional generalization, as monolithic policies do not decide which skill to reuse when scenes change; (ii) classical task-and-motion planning (TAMP) offers compositionality but has prohibitive planning latency, preventing real-time failure recovery. We introduce SymSkill, a unified learning framework that combines the benefits of IL and TAMP, allowing compositional generalization and failure recovery in real-time. Offline, SymSkill jointly learns predicates, operators, and skills directly from unlabeled and unsegmented demonstrations. At execution time, upon specifying a conjunction of one or more learned predicates, SymSkill uses a symbolic planner to compose and reorder learned skills to achieve the symbolic goals, while performing recovery at both the motion and symbolic levels in real time. Coupled with a compliant controller, SymSkill enables safe and uninterrupted execution under human and environmental disturbances. In RoboCasa simulation, SymSkill can execute 12 single-step tasks with 85% success rate. Without additional data, it composes these skills into multi-step plans requiring up to 6 skill recompositions, recovering robustly from execution failures. On a real Franka robot, we demonstrate SymSkill, learning from 5 minutes of unsegmented and unlabeled play data, is capable of performing multiple tasks simply by goal specifications. The source code and additional analysis can be found on https://sites.google.com/view/symskill.

Robot navigation in large, complex, and unknown indoor environments is a challenging problem. The existing approaches, such as traditional sampling-based methods, struggle with resolution control and scalability, while imitation learning-based methods require a large amount of demonstration data. Active Neural Time Fields (ANTFields) have recently emerged as a promising solution by using local observations to learn cost-to-go functions without relying on demonstrations. Despite their potential, these methods are hampered by challenges such as spectral bias and catastrophic forgetting, which diminish their effectiveness in complex scenarios. To address these issues, our approach decomposes the planning problem into a hierarchical structure. At the high level, a sparse graph captures the environment's global connectivity, while at the low level, a planner based on neural fields navigates local obstacles by solving the Eikonal PDE. This physics-informed strategy overcomes common pitfalls like spectral bias and neural field fitting difficulties, resulting in a smooth and precise representation of the cost landscape. We validate our framework in large-scale environments, demonstrating its enhanced adaptability and precision compared to previous methods, and highlighting its potential for online exploration, mapping, and real-world navigation.

Control Barrier Functions (CBFs) are a powerful tool for ensuring the safety of autonomous systems, yet applying them to nonholonomic robots in cluttered, dynamic environments remains an open challenge. State-of-the-art methods often rely on collision-cone or velocity-obstacle constraints which, by only considering the angle of the relative velocity, are inherently conservative and can render the CBF-based quadratic program infeasible, particularly in dense scenarios. To address this issue, we propose a Dynamic Parabolic Control Barrier Function (DPCBF) that defines the safe set using a parabolic boundary. The parabola's vertex and curvature dynamically adapt based on both the distance to an obstacle and the magnitude of the relative velocity, creating a less restrictive safety constraint. We prove that the proposed DPCBF is valid for a kinematic bicycle model subject to input constraints. Extensive comparative simulations demonstrate that our DPCBF-based controller significantly enhances navigation success rates and QP feasibility compared to baseline methods. Our approach successfully navigates through dense environments with up to 100 dynamic obstacles, scenarios where collision cone-based methods fail due to infeasibility.

Robots must adapt to diverse human instructions and operate safely in unstructured, open-world environments. Recent Vision-Language models (VLMs) offer strong priors for grounding language and perception, but remain difficult to steer for navigation due to differences in action spaces and pretraining objectives that hamper transferability to robotics tasks. Towards addressing this, we introduce VENTURA, a vision-language navigation system that finetunes internet-pretrained image diffusion models for path planning. Instead of directly predicting low-level actions, VENTURA generates a path mask (i.e. a visual plan) in image space that captures fine-grained, context-aware navigation behaviors. A lightweight behavior-cloning policy grounds these visual plans into executable trajectories, yielding an interface that follows natural language instructions to generate diverse robot behaviors. To scale training, we supervise on path masks derived from self-supervised tracking models paired with VLM-augmented captions, avoiding manual pixel-level annotation or highly engineered data collection setups. In extensive real-world evaluations, VENTURA outperforms state-of-the-art foundation model baselines on object reaching, obstacle avoidance, and terrain preference tasks, improving success rates by 33% and reducing collisions by 54% across both seen and unseen scenarios. Notably, we find that VENTURA generalizes to unseen combinations of distinct tasks, revealing emergent compositional capabilities. Videos, code, and additional materials: https://venturapath.github.io