Unifying Model Predictive Path Integral Control, Reinforcement Learning, and Diffusion Models for Optimal Control and Planning

Model Predictive Path Integral (MPPI) control, Reinforcement Learning (RL), and Diffusion Models have each demonstrated strong performance in trajectory optimization, decision-making, and motion planning. However, these approaches have traditionally been treated as distinct methodologies with separate optimization frameworks. In this work, we establish a unified perspective that connects MPPI, RL, and Diffusion Models through gradient-based optimization on the Gibbs measure. We first show that MPPI can be interpreted as performing gradient ascent on a smoothed energy function. We then demonstrate that Policy Gradient methods reduce to MPPI when treating policy parameters as control variables under a fixed initial state. Additionally, we establish that the reverse sampling process in diffusion models follows the same update rule as MPPI.

Efficient and Optimal Policy Gradient Algorithm for Corrupted Multi-armed Bandits

In this paper, we consider the stochastic multi-armed bandits problem with adversarial corruptions, where the random rewards of the arms are partially modified by an adversary to fool the algorithm. We apply the policy gradient algorithm SAMBA to this setting, and show that it is computationally efficient, and achieves a state-of-the-art $O(K\log T/\Delta) + O(C/\Delta)$ regret upper bound, where $K$ is the number of arms, $C$ is the unknown corruption level, $\Delta$ is the minimum expected reward gap between the best arm and other ones, and $T$ is the time horizon. Compared with the best existing efficient algorithm (e.g., CBARBAR), whose regret upper bound is $O(K\log^2 T/\Delta) + O(C)$, we show that SAMBA reduces one $\log T$ factor in the regret bound, while maintaining the corruption-dependent term to be linear with $C$. This is indeed asymptotically optimal. We also conduct simulations to demonstrate the effectiveness of SAMBA, and the results show that SAMBA outperforms existing baselines.

Accelerating Model-Based Reinforcement Learning with State-Space World Models

Reinforcement learning (RL) is a powerful approach for robot learning. However, model-free RL (MFRL) requires a large number of environment interactions to learn successful control policies. This is due to the noisy RL training updates and the complexity of robotic systems, which typically involve highly non-linear dynamics and noisy sensor signals. In contrast, model-based RL (MBRL) not only trains a policy but simultaneously learns a world model that captures the environment's dynamics and rewards. The world model can either be used for planning, for data collection, or to provide first-order policy gradients for training. Leveraging a world model significantly improves sample efficiency compared to model-free RL. However, training a world model alongside the policy increases the computational complexity, leading to longer training times that are often intractable for complex real-world scenarios. In this work, we propose a new method for accelerating model-based RL using state-space world models. Our approach leverages state-space models (SSMs) to parallelize the training of the dynamics model, which is typically the main computational bottleneck. Additionally, we propose an architecture that provides privileged information to the world model during training, which is particularly relevant for partially observable environments. We evaluate our method in several real-world agile quadrotor flight tasks, involving complex dynamics, for both fully and partially observable environments. We demonstrate a significant speedup, reducing the world model training time by up to 10 times, and the overall MBRL training time by up to 4 times. This benefit comes without compromising performance, as our method achieves similar sample efficiency and task rewards to state-of-the-art MBRL methods.

Neural Combinatorial Optimization via Preference Optimization

Neural Combinatorial Optimization (NCO) has emerged as a promising approach for NP-hard problems. However, prevailing RL-based methods suffer from low sample efficiency due to sparse rewards and underused solutions. We propose Preference Optimization for Combinatorial Optimization (POCO), a training paradigm that leverages solution preferences via objective values. It introduces: (1) an efficient preference pair construction for better explore and exploit solutions, and (2) a novel loss function that adaptively scales gradients via objective differences, removing reliance on reward models or reference policies. Experiments on Job-Shop Scheduling (JSP), Traveling Salesman (TSP), and Flexible Job-Shop Scheduling (FJSP) show POCO outperforms state-of-the-art neural methods, reducing optimality gaps impressively with efficient inference. POCO is architecture-agnostic, enabling seamless integration with existing NCO models, and establishes preference optimization as a principled framework for combinatorial optimization.

Reevaluating Policy Gradient Methods for Imperfect-Information Games

In the past decade, motivated by the putative failure of naive self-play deep reinforcement learning (DRL) in adversarial imperfect-information games, researchers have developed numerous DRL algorithms based on fictitious play (FP), double oracle (DO), and counterfactual regret minimization (CFR). In light of recent results of the magnetic mirror descent algorithm, we hypothesize that simpler generic policy gradient methods like PPO are competitive with or superior to these FP, DO, and CFR-based DRL approaches. To facilitate the resolution of this hypothesis, we implement and release the first broadly accessible exact exploitability computations for four large games. Using these games, we conduct the largest-ever exploitability comparison of DRL algorithms for imperfect-information games. Over 5600 training runs, FP, DO, and CFR-based approaches fail to outperform generic policy gradient methods. Code is available at https://github.com/nathanlct/IIG-RL-Benchmark and https://github.com/gabrfarina/exp-a-spiel .

Deep Reinforcement Learning based Autonomous Decision-Making for Cooperative UAVs: A Search and Rescue Real World Application

This paper proposes a holistic framework for autonomous guidance, navigation, and task distribution among multi-drone systems operating in Global Navigation Satellite System (GNSS)-denied indoor settings. We advocate for a Deep Reinforcement Learning (DRL)-based guidance mechanism, utilising the Twin Delayed Deep Deterministic Policy Gradient algorithm. To improve the efficiency of the training process, we incorporate an Artificial Potential Field (APF)-based reward structure, enabling the agent to refine its movements, thereby promoting smoother paths and enhanced obstacle avoidance in indoor contexts. Furthermore, we tackle the issue of task distribution among cooperative UAVs through a DRL-trained Graph Convolutional Network (GCN). This GCN represents the interactions between drones and tasks, facilitating dynamic and real-time task allocation that reflects the current environmental conditions and the capabilities of the drones. Such an approach fosters effective coordination and collaboration among multiple drones during search and rescue operations or other exploratory endeavours. Lastly, to ensure precise odometry in environments lacking GNSS, we employ Light Detection And Ranging Simultaneous Localisation and Mapping complemented by a depth camera to mitigate the hallway problem. This integration offers robust localisation and mapping functionalities, thereby enhancing the systems dependability in indoor navigation. The proposed multi-drone framework not only elevates individual navigation capabilities but also optimises coordinated task allocation in complex, obstacle-laden environments. Experimental evaluations conducted in a setup tailored to meet the requirements of the NATO Sapience Autonomous Cooperative Drone Competition demonstrate the efficacy of the proposed system, yielding outstanding results and culminating in a first-place finish in the 2024 Sapience competition.

REINFORCE Adversarial Attacks on Large Language Models: An Adaptive, Distributional, and Semantic Objective

To circumvent the alignment of large language models (LLMs), current optimization-based adversarial attacks usually craft adversarial prompts by maximizing the likelihood of a so-called affirmative response. An affirmative response is a manually designed start of a harmful answer to an inappropriate request. While it is often easy to craft prompts that yield a substantial likelihood for the affirmative response, the attacked model frequently does not complete the response in a harmful manner. Moreover, the affirmative objective is usually not adapted to model-specific preferences and essentially ignores the fact that LLMs output a distribution over responses. If low attack success under such an objective is taken as a measure of robustness, the true robustness might be grossly overestimated. To alleviate these flaws, we propose an adaptive and semantic optimization problem over the population of responses. We derive a generally applicable objective via the REINFORCE policy-gradient formalism and demonstrate its efficacy with the state-of-the-art jailbreak algorithms Greedy Coordinate Gradient (GCG) and Projected Gradient Descent (PGD). For example, our objective doubles the attack success rate (ASR) on Llama3 and increases the ASR from 2% to 50% with circuit breaker defense.

A Multi-Agent DRL-Based Framework for Optimal Resource Allocation and Twin Migration in the Multi-Tier Vehicular Metaverse

Although multi-tier vehicular Metaverse promises to transform vehicles into essential nodes -- within an interconnected digital ecosystem -- using efficient resource allocation and seamless vehicular twin (VT) migration, this can hardly be achieved by the existing techniques operating in a highly dynamic vehicular environment, since they can hardly balance multi-objective optimization problems such as latency reduction, resource utilization, and user experience (UX). To address these challenges, we introduce a novel multi-tier resource allocation and VT migration framework that integrates Graph Convolutional Networks (GCNs), a hierarchical Stackelberg game-based incentive mechanism, and Multi-Agent Deep Reinforcement Learning (MADRL). The GCN-based model captures both spatial and temporal dependencies within the vehicular network; the Stackelberg game-based incentive mechanism fosters cooperation between vehicles and infrastructure; and the MADRL algorithm jointly optimizes resource allocation and VT migration in real time. By modeling this dynamic and multi-tier vehicular Metaverse as a Markov Decision Process (MDP), we develop a MADRL-based algorithm dubbed the Multi-Objective Multi-Agent Deep Deterministic Policy Gradient (MO-MADDPG), which can effectively balances the various conflicting objectives. Extensive simulations validate the effectiveness of this algorithm that is demonstrated to enhance scalability, reliability, and efficiency while considerably improving latency, resource utilization, migration cost, and overall UX by 12.8%, 9.7%, 14.2%, and 16.1%, respectively.

MoE-Loco: Mixture of Experts for Multitask Locomotion

We present MoE-Loco, a Mixture of Experts (MoE) framework for multitask locomotion for legged robots. Our method enables a single policy to handle diverse terrains, including bars, pits, stairs, slopes, and baffles, while supporting quadrupedal and bipedal gaits. Using MoE, we mitigate the gradient conflicts that typically arise in multitask reinforcement learning, improving both training efficiency and performance. Our experiments demonstrate that different experts naturally specialize in distinct locomotion behaviors, which can be leveraged for task migration and skill composition. We further validate our approach in both simulation and real-world deployment, showcasing its robustness and adaptability.

A Reinforcement Learning Approach to Non-prehensile Manipulation through Sliding

Although robotic applications increasingly demand versatile and dynamic object handling, most existing techniques are predominantly focused on grasp-based manipulation, limiting their applicability in non-prehensile tasks. To address this need, this study introduces a Deep Deterministic Policy Gradient (DDPG) reinforcement learning framework for efficient non-prehensile manipulation, specifically for sliding an object on a surface. The algorithm generates a linear trajectory by precisely controlling the acceleration of a robotic arm rigidly coupled to the horizontal surface, enabling the relative manipulation of an object as it slides on top of the surface. Furthermore, two distinct algorithms have been developed to estimate the frictional forces dynamically during the sliding process. These algorithms provide online friction estimates after each action, which are fed back into the actor model as critical feedback after each action. This feedback mechanism enhances the policy's adaptability and robustness, ensuring more precise control of the platform's acceleration in response to varying surface condition. The proposed algorithm is validated through simulations and real-world experiments. Results demonstrate that the proposed framework effectively generalizes sliding manipulation across varying distances and, more importantly, adapts to different surfaces with diverse frictional properties. Notably, the trained model exhibits zero-shot sim-to-real transfer capabilities.