ABSTRACT Multi - agent reinforcement learning (MARL) is increasingly ubiquitous in training dynamic and adaptive synthetic characters for interactive simulations on geo - specific terrains. Frameworks such as Unity's ML - Agents help to make such reinforcement learning e xperiments more accessible to the simulation community. Military training simulations also benefit from advances in MARL, but they have immense computational requirements due to their complex, continuous, stochastic, partially observable, non - stationary, a nd doctrine - based nature. Furthermore, these simulations require geo - specific terrains, further exacerbating the computational resources problem. In our research, we leverage Unity's waypoints to automatically generate multi - layered representation abstract ions of the geo - specific terrains to scale up reinforcement learning while still allowing the transfer of learned policies between different representations. Our early exploratory results on a novel MARL scenario, where each side has differing objectives, indicate that waypoint - based navigation enables faster and more efficient learning while producing trajectories similar to those taken by expert human players in CSGO gaming environments. This research points out the potential of waypoint - based navigation for reducing the computational costs of developing and training MARL models for military training simulations, where geo - specific terrains and differing objectives are crucial. ABOUT THE AUTHORS Volkan Ustun is the Associate Director of the Human - Inspired Adaptive Teaming Systems Group at the USC I nstitute for Creative Technologies .

Multi-agent reinforcement learning (MARL) faces two critical bottlenecks distinct from single-agent RL: credit assignment in cooperative tasks and partial observability of environmental states. We propose LERO, a framework integrating Large language models (LLMs) with evolutionary optimization to address these MARL-specific challenges. The solution centers on two LLM-generated components: a hybrid reward function that dynamically allocates individual credit through reward decomposition, and an observation enhancement function that augments partial observations with inferred environmental context. An evolutionary algorithm optimizes these components through iterative MARL training cycles, where top-performing candidates guide subsequent LLM generations. Evaluations in Multi-Agent Particle Environments (MPE) demonstrate LERO's superiority over baseline methods, with improved task performance and training efficiency.

Recent developments in deep reinforcement learning have been very successful in learning complex, previously intractable problems. Sample efficiency and local optimality, however, remain significant challenges. To address these challenges, novelty-driven exploration strategies have emerged and shown promising potential. Unfortunately, no single algorithm outperforms all others in all tasks and most of them struggle with tasks with high-dimensional and complex observations. In this work, we propose Adventurer, a novelty-driven exploration algorithm that is based on Bidirectional Generative Adversarial Networks (BiGAN), where BiGAN is trained to estimate state novelty. Intuitively, a generator that has been trained on the distribution of visited states should only be able to generate a state coming from the distribution of visited states. As a result, novel states using the generator to reconstruct input states from certain latent representations would lead to larger reconstruction errors. We show that BiGAN performs well in estimating state novelty for complex observations. This novelty estimation method can be combined with intrinsic-reward-based exploration. Our empirical results show that Adventurer produces competitive results on a range of popular benchmark tasks, including continuous robotic manipulation tasks (e.g. Mujoco robotics) and high-dimensional image-based tasks (e.g. Atari games).

--Mining process optimization, particularly truck dispatch scheduling, is a critical factor in enhancing the efficiency of open-pit mining operations. However, the dynamic and stochastic nature of mining environments--characterized by uncertainties such as equipment failures, truck maintenance, and variable haul cycle times--poses significant challenges for traditional optimization methods. While Reinforcement Learning (RL) has demonstrated promise in adaptive decision-making for mining logistics, its practical deployment requires rigorous evaluation in realistic and customizable simulation environments. T o address this challenge, we introduce Mining-Gym, a configurable, open-source benchmarking environment designed for training, testing, and comparing RL algorithms in mining process optimization. Built on Discrete Event Simulation (DES) and seamlessly integrated with the OpenAI Gym interface, Mining-Gym offers a structured testbed that enables the direct application of advanced RL algorithms from Stable Baselines. The framework models key mining-specific uncertainties, such as equipment failures, queue congestion, and stochasticity of mining processes, ensuring a realistic and adaptive learning environment. Additionally, a graphic user interface (GUI) for easy parameter selection for mine-site configuration, comprehensive data logging system, a built-in KPI dashboard and real-time representative visualization of mine-site enables in-depth performance analysis, facilitating standardized, reproducible evaluation across multiple RL strategies and baseline heuristics. INING process optimization aims to enhance efficiency and productivity by improving resource allocation, equipment scheduling, and material handling. However, these operations are highly complex, influenced by dynamic factors such as equipment failures, fluctuating ore quality, and unpredictable environmental conditions. Traditional optimization methods, such as linear programming and heuristics, struggle to adapt in real time, leading to inefficiencies and increased costs.

Offline reinforcement learning (RL) aims to learn from historical data without requiring (costly) access to the environment. To facilitate offline RL research, we previously introduced NeoRL, which highlighted that datasets from real-world tasks are often conservative and limited. With years of experience applying offline RL to various domains, we have identified additional real-world challenges. These include extremely conservative data distributions produced by deployed control systems, delayed action effects caused by high-latency transitions, external factors arising from the uncontrollable variance of transitions, and global safety constraints that are difficult to evaluate during the decision-making process. These challenges are underrepresented in previous benchmarks but frequently occur in real-world tasks. To address this, we constructed the extended Near Real-World Offline RL Benchmark (NeoRL-2), which consists of 7 datasets from 7 simulated tasks along with their corresponding evaluation simulators. Benchmarking results from state-of-the-art offline RL approaches demonstrate that current methods often struggle to outperform the data-collection behavior policy, highlighting the need for more effective methods. We hope NeoRL-2 will accelerate the development of reinforcement learning algorithms for real-world applications. The benchmark project page is available at https://github.com/polixir/NeoRL2.

Despite its extreme sample inefficiency, on-policy reinforcement learning has become a fundamental tool in real-world applications. With recent advances in GPU-driven simulation, the ability to collect vast amounts of data for RL training has scaled exponentially. However, studies show that current on-policy methods, such as PPO, fail to fully leverage the benefits of parallelized environments, leading to performance saturation beyond a certain scale. In contrast, Evolutionary Algorithms (EAs) excel at increasing diversity through randomization, making them a natural complement to RL. However, existing EvoRL methods have struggled to gain widespread adoption due to their extreme sample inefficiency. To address these challenges, we introduce Evolutionary Policy Optimization (EPO), a novel policy gradient algorithm that combines the strengths of EA and policy gradients. We show that EPO significantly improves performance across diverse and challenging environments, demonstrating superior scalability with parallelized simulations.

Without using explicit reward, direct preference optimization (DPO) employs paired human preference data to fine-tune generative models, a method that has garnered considerable attention in large language models (LLMs). However, exploration of aligning text-to-image (T2I) diffusion models with human preferences remains limited. In comparison to supervised fine-tuning, existing methods that align diffusion model suffer from low training efficiency and subpar generation quality due to the long Markov chain process and the intractability of the reverse process. To address these limitations, we introduce DDIM-InPO, an efficient method for direct preference alignment of diffusion models. Our approach conceptualizes diffusion model as a single-step generative model, allowing us to fine-tune the outputs of specific latent variables selectively. In order to accomplish this objective, we first assign implicit rewards to any latent variable directly via a reparameterization technique. Then we construct an Inversion technique to estimate appropriate latent variables for preference optimization. This modification process enables the diffusion model to only fine-tune the outputs of latent variables that have a strong correlation with the preference dataset. Experimental results indicate that our DDIM-InPO achieves state-of-the-art performance with just 400 steps of fine-tuning, surpassing all preference aligning baselines for T2I diffusion models in human preference evaluation tasks.

Robotic surgery is a rapidly developing field that can greatly benefit from the automation of surgical tasks. However, training techniques such as Reinforcement Learning (RL) require a high number of task repetitions, which are generally unsafe and impractical to perform on real surgical systems. This stresses the need for simulated surgical environments, which are not only realistic, but also computationally efficient and scalable. We introduce FF-SRL (Fast and Flexible Surgical Reinforcement Learning), a high-performance learning environment for robotic surgery. In FF-SRL both physics simulation and RL policy training reside entirely on a single GPU. This avoids typical bottlenecks associated with data transfer between the CPU and GPU, leading to accelerated learning rates. Our results show that FF-SRL reduces the training time of a complex tissue manipulation task by an order of magnitude, down to a couple of minutes, compared to a common CPU/GPU simulator. Such speed-up may facilitate the experimentation with RL techniques and contribute to the development of new generation of surgical systems. To this end, we make our code publicly available to the community.

Reinforcement learning in non-stationary environments is challenging due to abrupt and unpredictable changes in dynamics, often causing traditional algorithms to fail to converge. However, in many real-world cases, non-stationarity has some structure that can be exploited to develop algorithms and facilitate theoretical analysis. We introduce one such structure, Switching Non-Stationary Markov Decision Processes (SNS-MDP), where environments switch over time-based on an underlying Markov chain. Under a fixed policy, the value function of an SNS-MDP admits a closed-form solution determined by the Markov chain's statistical properties, and despite the inherent non-stationarity, Temporal Difference (TD) learning methods still converge to the correct value function. Furthermore, policy improvement can be performed, and it is shown that policy iteration converges to the optimal policy. Moreover, since Q-learning converges to the optimal Q-function, it likewise yields the corresponding optimal policy. To illustrate the practical advantages of SNS-MDPs, we present an example in communication networks where channel noise follows a Markovian pattern, demonstrating how this framework can effectively guide decision-making in complex, time-varying contexts.

This work addresses the challenge of personalizing trajectories generated in automated decision-making systems by introducing a resource-efficient approach that enables rapid adaptation to individual users' preferences. Our method leverages a pretrained conditional diffusion model with Preference Latent Embeddings (PLE), trained on a large, reward-free offline dataset. The PLE serves as a compact representation for capturing specific user preferences. By adapting the pretrained model using our proposed preference inversion method, which directly optimizes the learnable PLE, we achieve superior alignment with human preferences compared to existing solutions like Reinforcement Learning from Human Feedback (RLHF) and Low-Rank Adaptation (LoRA). To better reflect practical applications, we create a benchmark experiment using real human preferences on diverse, high-reward trajectories.