Online Preference-based Reinforcement Learning with Self-augmented Feedback from Large Language Model

Preference-based reinforcement learning (PbRL) provides a powerful paradigm to avoid meticulous reward engineering by learning rewards based on human preferences. However, real-time human feedback is hard to obtain in online tasks. Most work suppose there is a "scripted teacher" that utilizes privileged predefined reward to provide preference feedback. In this paper, we propose a RL Self-augmented Large Language Model Feedback (RL-SaLLM-F) technique that does not rely on privileged information for online PbRL. RL-SaLLM-F leverages the reflective and discriminative capabilities of LLM to generate self-augmented trajectories and provide preference labels for reward learning. First, we identify an failure issue in LLM-based preference discrimination, specifically "query ambiguity", in online PbRL. Then LLM is employed to provide preference labels and generate self-augmented imagined trajectories that better achieve the task goal, thereby enhancing the quality and efficiency of feedback. Additionally, a double-check mechanism is introduced to mitigate randomness in the preference labels, improving the reliability of LLM feedback. The experiment across multiple tasks in the MetaWorld benchmark demonstrates the specific contributions of each proposed module in RL-SaLLM-F, and shows that self-augmented LLM feedback can effectively replace the impractical "scripted teacher" feedback. In summary, RL-SaLLM-F introduces a new direction of feedback acquisition in online PbRL that does not rely on any online privileged information, offering an efficient and lightweight solution with LLM-driven feedback.

In-Dataset Trajectory Return Regularization for Offline Preference-based Reinforcement Learning

Offline preference-based reinforcement learning (PbRL) typically operates in two phases: first, use human preferences to learn a reward model and annotate rewards for a reward-free offline dataset; second, learn a policy by optimizing the learned reward via offline RL. However, accurately modeling step-wise rewards from trajectory-level preference feedback presents inherent challenges. The reward bias introduced, particularly the overestimation of predicted rewards, leads to optimistic trajectory stitching, which undermines the pessimism mechanism critical to the offline RL phase. To address this challenge, we propose In-Dataset Trajectory Return Regularization (DTR) for offline PbRL, which leverages conditional sequence modeling to mitigate the risk of learning inaccurate trajectory stitching under reward bias. Specifically, DTR employs Decision Transformer and TD-Learning to strike a balance between maintaining fidelity to the behavior policy with high in-dataset trajectory returns and selecting optimal actions based on high reward labels. Additionally, we introduce an ensemble normalization technique that effectively integrates multiple reward models, balancing the tradeoff between reward differentiation and accuracy. Empirical evaluations on various benchmarks demonstrate the superiority of DTR over other state-of-the-art baselines

Preliminary Investigation into Data Scaling Laws for Imitation Learning-Based End-to-End Autonomous Driving

The end-to-end autonomous driving paradigm has recently attracted lots of attention due to its scalability. However, existing methods are constrained by the limited scale of real-world data, which hinders a comprehensive exploration of the scaling laws associated with end-to-end autonomous driving. To address this issue, we collected substantial data from various driving scenarios and behaviors and conducted an extensive study on the scaling laws of existing imitation learning-based end-to-end autonomous driving paradigms. Specifically, approximately 4 million demonstrations from 23 different scenario types were gathered, amounting to over 30,000 hours of driving demonstrations. We performed open-loop evaluations and closed-loop simulation evaluations in 1,400 diverse driving demonstrations (1,300 for open-loop and 100 for closed-loop) under stringent assessment conditions. Through experimental analysis, we discovered that (1) the performance of the driving model exhibits a power-law relationship with the amount of training data; (2) a small increase in the quantity of long-tailed data can significantly improve the performance for the corresponding scenarios; (3) appropriate scaling of data enables the model to achieve combinatorial generalization in novel scenes and actions. Our results highlight the critical role of data scaling in improving the generalizability of models across diverse autonomous driving scenarios, assuring safe deployment in the real world. Project repository: https://github.com/ucaszyp/Driving-Scaling-Law

Meta-DT: Offline Meta-RL as Conditional Sequence Modeling with World Model Disentanglement

A longstanding goal of artificial general intelligence is highly capable generalists that can learn from diverse experiences and generalize to unseen tasks. The language and vision communities have seen remarkable progress toward this trend by scaling up transformer-based models trained on massive datasets, while reinforcement learning (RL) agents still suffer from poor generalization capacity under such paradigms. To tackle this challenge, we propose Meta Decision Transformer (Meta-DT), which leverages the sequential modeling ability of the transformer architecture and robust task representation learning via world model disentanglement to achieve efficient generalization in offline meta-RL. We pretrain a context-aware world model to learn a compact task representation, and inject it as a contextual condition to the causal transformer to guide task-oriented sequence generation. Then, we subtly utilize history trajectories generated by the meta-policy as a self-guided prompt to exploit the architectural inductive bias. We select the trajectory segment that yields the largest prediction error on the pretrained world model to construct the prompt, aiming to encode task-specific information complementary to the world model maximally. Notably, the proposed framework eliminates the requirement of any expert demonstration or domain knowledge at test time. Experimental results on MuJoCo and Meta-World benchmarks across various dataset types show that Meta-DT exhibits superior few and zero-shot generalization capacity compared to strong baselines while being more practical with fewer prerequisites. Our code is available at https://github.com/NJU-RL/Meta-DT.

SELU: Self-Learning Embodied MLLMs in Unknown Environments

Recently, multimodal large language models (MLLMs) have demonstrated strong visual understanding and decision-making capabilities, enabling the exploration of autonomously improving MLLMs in unknown environments. However, external feedback like human or environmental feedback is not always available. To address this challenge, existing methods primarily focus on enhancing the decision-making capabilities of MLLMs through voting and scoring mechanisms, while little effort has been paid to improving the environmental comprehension of MLLMs in unknown environments. To fully unleash the self-learning potential of MLLMs, we propose a novel actor-critic self-learning paradigm, dubbed SELU, inspired by the actor-critic paradigm in reinforcement learning. The critic employs self-asking and hindsight relabeling to extract knowledge from interaction trajectories collected by the actor, thereby augmenting its environmental comprehension. Simultaneously, the actor is improved by the self-feedback provided by the critic, enhancing its decision-making. We evaluate our method in the AI2-THOR and VirtualHome environments, and SELU achieves critic improvements of approximately 28% and 30%, and actor improvements of about 20% and 24% via self-learning.

Discretizing Continuous Action Space with Unimodal Probability Distributions for On-Policy Reinforcement Learning

For on-policy reinforcement learning, discretizing action space for continuous control can easily express multiple modes and is straightforward to optimize. However, without considering the inherent ordering between the discrete atomic actions, the explosion in the number of discrete actions can possess undesired properties and induce a higher variance for the policy gradient estimator. In this paper, we introduce a straightforward architecture that addresses this issue by constraining the discrete policy to be unimodal using Poisson probability distributions. This unimodal architecture can better leverage the continuity in the underlying continuous action space using explicit unimodal probability distributions. We conduct extensive experiments to show that the discrete policy with the unimodal probability distribution provides significantly faster convergence and higher performance for on-policy reinforcement learning algorithms in challenging control tasks, especially in highly complex tasks such as Humanoid. We provide theoretical analysis on the variance of the policy gradient estimator, which suggests that our attentively designed unimodal discrete policy can retain a lower variance and yield a stable learning process.

PlanAgent: A Multi-modal Large Language Agent for Closed-loop Vehicle Motion Planning

Vehicle motion planning is an essential component of autonomous driving technology. Current rule-based vehicle motion planning methods perform satisfactorily in common scenarios but struggle to generalize to long-tailed situations. Meanwhile, learning-based methods have yet to achieve superior performance over rule-based approaches in large-scale closed-loop scenarios. To address these issues, we propose PlanAgent, the first mid-to-mid planning system based on a Multi-modal Large Language Model (MLLM). MLLM is used as a cognitive agent to introduce human-like knowledge, interpretability, and common-sense reasoning into the closed-loop planning. Specifically, PlanAgent leverages the power of MLLM through three core modules. First, an Environment Transformation module constructs a Bird's Eye View (BEV) map and a lane-graph-based textual description from the environment as inputs. Second, a Reasoning Engine module introduces a hierarchical chain-of-thought from scene understanding to lateral and longitudinal motion instructions, culminating in planner code generation. Last, a Reflection module is integrated to simulate and evaluate the generated planner for reducing MLLM's uncertainty. PlanAgent is endowed with the common-sense reasoning and generalization capability of MLLM, which empowers it to effectively tackle both common and complex long-tailed scenarios. Our proposed PlanAgent is evaluated on the large-scale and challenging nuPlan benchmarks. A comprehensive set of experiments convincingly demonstrates that PlanAgent outperforms the existing state-of-the-art in the closed-loop motion planning task. Codes will be soon released.

Learning Future Representation with Synthetic Observations for Sample-efficient Reinforcement Learning

In visual Reinforcement Learning (RL), upstream representation learning largely determines the effect of downstream policy learning. Employing auxiliary tasks allows the agent to enhance visual representation in a targeted manner, thereby improving the sample efficiency and performance of downstream RL. Prior advanced auxiliary tasks all focus on how to extract as much information as possible from limited experience (including observations, actions, and rewards) through their different auxiliary objectives, whereas in this article, we first start from another perspective: auxiliary training data. We try to improve auxiliary representation learning for RL by enriching auxiliary training data, proposing \textbf{L}earning \textbf{F}uture representation with \textbf{S}ynthetic observations \textbf{(LFS)}, a novel self-supervised RL approach. Specifically, we propose a training-free method to synthesize observations that may contain future information, as well as a data selection approach to eliminate unqualified synthetic noise. The remaining synthetic observations and real observations then serve as the auxiliary data to achieve a clustering-based temporal association task for representation learning. LFS allows the agent to access and learn observations that have not yet appeared in advance, so as to quickly understand and exploit them when they occur later. In addition, LFS does not rely on rewards or actions, which means it has a wider scope of application (e.g., learning from video) than recent advanced auxiliary tasks. Extensive experiments demonstrate that our LFS exhibits state-of-the-art RL sample efficiency on challenging continuous control and enables advanced visual pre-training based on action-free video demonstrations.

Advancing Object Goal Navigation Through LLM-enhanced Object Affinities Transfer

In object goal navigation, agents navigate towards objects identified by category labels using visual and spatial information. Previously, solely network-based methods typically rely on historical data for object affinities estimation, lacking adaptability to new environments and unseen targets. Simultaneously, employing Large Language Models (LLMs) for navigation as either planners or agents, though offering a broad knowledge base, is cost-inefficient and lacks targeted historical experience. Addressing these challenges, we present the LLM-enhanced Object Affinities Transfer (LOAT) framework, integrating LLM-derived object semantics with network-based approaches to leverage experiential object affinities, thus improving adaptability in unfamiliar settings. LOAT employs a dual-module strategy: a generalized affinities module for accessing LLMs' vast knowledge and an experiential affinities module for applying learned object semantic relationships, complemented by a dynamic fusion module harmonizing these information sources based on temporal context. The resulting scores activate semantic maps before feeding into downstream policies, enhancing navigation systems with context-aware inputs. Our evaluations in AI2-THOR and Habitat simulators demonstrate improvements in both navigation success rates and efficiency, validating the LOAT's efficacy in integrating LLM insights for improved object goal navigation.

FM3Q: Factorized Multi-Agent MiniMax Q-Learning for Two-Team Zero-Sum Markov Game

Many real-world applications involve some agents that fall into two teams, with payoffs that are equal within the same team but of opposite sign across the opponent team. The so-called two-team zero-sum Markov games (2t0sMGs) can be resolved with reinforcement learning in recent years. However, existing methods are thus inefficient in light of insufficient consideration of intra-team credit assignment, data utilization and computational intractability. In this paper, we propose the individual-global-minimax (IGMM) principle to ensure the coherence between two-team minimax behaviors and the individual greedy behaviors through Q functions in 2t0sMGs. Based on it, we present a novel multi-agent reinforcement learning framework, Factorized Multi-Agent MiniMax Q-Learning (FM3Q), which can factorize the joint minimax Q function into individual ones and iteratively solve for the IGMM-satisfied minimax Q functions for 2t0sMGs. Moreover, an online learning algorithm with neural networks is proposed to implement FM3Q and obtain the deterministic and decentralized minimax policies for two-team players. A theoretical analysis is provided to prove the convergence of FM3Q. Empirically, we use three environments to evaluate the learning efficiency and final performance of FM3Q and show its superiority on 2t0sMGs.