Jul 2026· International Journal of Innovative Computing· Vol 16, pp. 123-131· 0 citations· 20 references
TL;DR
The findings demonstrate MARL’s promise in solving navigation problems efficiently and provide concrete recommendations for tuning training parameters and network structures to enhance performance and robustness.
Abstract
This study explores the potential of Multiagent Reinforcement Learning (MARL) for autonomous navigation in a discrete two-dimensional environment. We design and implement an agent that learns an optimal path through a 5×5 grid via repeated interactions and a reward-based mechanism. Over 500 training episodes, we examine the convergence speed of the learned policy, the stability of agent behavior, and the success rate in reaching the goal. Our approach combines artificial neural networks with a multiagent framework, enabling decentralized decision making and scalable adaptation. We discuss critical factors affecting learning stability, including reward function design and network architecture, and outline avenues for extending the methodology to more complex, real-time tasks. The findings demonstrate MARL’s promise in solving navigation problems efficiently and provide concrete recommendations for tuning training parameters and network structures to enhance performance and robustness.
The performance and generalization of deep reinforcement learning (RL) agents are strongly influenced by the structure of their training environments, yet environment design is typically manual and rarely optimized in a principled manner. Among them, maze-solving tasks present several challenges for evaluating reinforcement learning methods in navigation and sequential decision-making. In this paper, we propose a Maze Evolution framework, denoted as ME, where an evolutionary algorithm (EA) searches a constrained space of grid-based mazes to produce a best evolved environment set for RL with desirable properties (e.g., feasibility and controlled structural difficulty). The resulting evolved environment set is then used to train a Deep Q-Network (DQN) agent. The performance of ME is evaluated on a test environment set and compared against RL agents that is trained on a fixed environment. Experimental results demonstrates that our method can increase the success rate of RL agent by 12% to achieve its goal.
Yang Huang, P. Harn· Proceedings of the Genetic a...· 0 citations
Reinforcement learning (RL) provides a framework for sequential decision making under explicit objectives. In its classical form, RL studies how an agent should act to maximise long-term reward in a dynamic environment. In richer settings, the problem extends beyond a single agent and fixed environment: intelligent behavior may require strategic interaction, adaptation to uncertainty, and reasoning over high-dimensional worlds. This thesis studies RL from two perspectives: algorithms in games and RL in the era of foundation models. The first part focuses on multi-agent RL in games. It examines how incentives, policies, and equilibrium concepts interact in competitive and general-sum environments, spanning two-player zero-sum games, large-scale video games, and multi-player settings with general structure. These works investigate learning in multi-agent systems and the behavior of RL methods in interactive environments. The second part studies RL with generative and foundation models, motivated by the idea that prior knowledge can enrich sequential decision making. Pretrained generative models and learned world models serve as representation tools and structured priors for planning, control, and policy optimization. The thesis develops diffusion-based world models, investigates RL for efficient video generation, explores generative models as policy classes, and studies interactive video world models in which actions shape future observations. It also addresses long-horizon modeling through architectures with memory. Together, these contributions present a unified view of RL as objective-driven adaptation in complex sequential domains. From strategic games to generative world models, the thesis highlights how RL connects decision making, environment modeling, and emerging foundation-model capabilities, offering a broader perspective on the principles underlying intelligent behavior.
Reinforcement learning (RL) is a method of training artificial intelligence agents to make decisions
through trial and error, rewarding good behavior and penalizing bad behavior until the agent learns an
effective strategy. This study compares three widely used RL algorithms for continuous robotic control:
Proximal Policy Optimization (PPO), which learns by making small, cautious updates to avoid unstable
training; Soft Actor-Critic (SAC), which encourages broad exploration by rewarding the agent for trying
diverse strategies alongside completing the task; and Advantage Actor-Critic (A2C), which uses two
neural networks simultaneously - one to decide actions and one to evaluate them. Three experimental
conditions were evaluated across three robotic simulation environments of increasing difficulty: standard
task performance, energy-aware performance (where agents were equally penalized for excessive energy
use alongside task completion), and transfer learning (where agents pre-trained on standard rewards were
fine-tuned on energy-aware rewards). SAC consistently outperformed PPO and A2C on dense-reward
environments under both standard and energy-constrained conditions. Effect sizes indicated some nonsignificant
differences were nonetheless practically large. Most strikingly, A2C exhibited highly unstable
performance under energy-aware rewards on MountainCarContinuous-v0, achieving a mean reward of
-18,231 compared to near-zero values for PPO and SAC, suggesting that A2C may require additional
reward shaping or tuning before use in energy-constrained robotic applications. Transfer learning
experiments revealed that pre-training on standard rewards generally hurt rather than helped adaptation
to energy-aware objectives. These findings offer practical guidelines for algorithm selection in energyconstrained
robotic systems such as prosthetic hands and autonomous vehicles.
Om Herur· American Journal of Student...· 0 citations
Multi-agent Reinforcement learning has gained significant attention for solving decision-making problems involving multiple autonomous agents. However, effective learning in MARL is still difficult due to environments, dependencies between agents, and poor exploration strategies. Although adaptive exploration and curriculum learning methods, such as Reward Prediction Error Adaptive Learning (RPEAL) along with Reward-Shaped Adaptive Curriculum Learning (RSACL), have produced good outcomes in single-agent reinforcement learning, their use in multi-agent contexts has not been thoroughly investigated. In this research, RPEAL and RSACL are introduced. This paper extends the previous single-agent work to the broader realm of cooperative multi-agent reinforcement learning. The introduced adaptive control mechanism are integrated into several popular multi-agent algorithms such as IPPO, CPPO, MADDPG, and MASAC are empirically compared in a standard petting-zoo environments. The experimental evaluation shows gains in these algorithms upon the introduction of adaptive control mechanisms, where the centralized critic outperforms the individual learners in terms of stability and convergence. Unlike previous works, which only considered single-agent reinforcement learning, in this paper we extend the RPEAL and RSACL to the multi-agent domain. To be specific, we propose team reward prediction error modeling with a centralized critic, as well as performance-driven curriculum learning for multi-agents.
B. Adwaith, Kevin Francis, Remya Nair T· International Conference on...· 0 citations
Reinforcement learning (RL) research has demonstrated success in both physical and simulated domains; however, the predominant methodology remains rooted in simulations. The predominance of simulations makes translating research to physical reality uncertain for both algorithms and researchers. We propose a physical platform that is designed to simplify the transition. In this paper, we present the Open Ant: a physical variant of the commonly used Gymnasium Ant environment, along with a simulation. We demonstrate that competent walking policies can be learned from scratch in approximately one hour directly from the physical robot's experience for two substantially different RL algorithms: SARSA($\lambda$) and Soft Actor-Critic (SAC). Separately, we show policies that were learned in simulation transfer to reality. We also examine how well the platform supports a nimble experimental ecosystem. Specifically, we observe the speed with which new users from diverse backgrounds achieve their first success with the platform, and how easily the platform can be repaired and updated when hardware issues arise. Both the hardware design and software are available as open-source on GitHub for ease of customization. In summary, we advocate for the use of the Open Ant for RL researchers who frequently use simulated environments, so they can more easily include robot experiments in their evaluations.
E. Lupu, Patrick Spieler, Khurram Javed et al.· 0 citations
The problem of moving from a starting point to a destination through a maze of walls and passages, known as maze solving, has been an active field of research in mathematics and computer science for decades with applications to robotics, logistics, network routing and game AI. Classical search algorithms such as Depth First Search (DFS), Breadth First Search (BFS) and A* provide deterministic guarantees but require complete prior knowledge of the environment and fail in dynamic or partially observable settings. Reinforcement learning (RL) provides a completely different paradigm. Agents learn the optimal navigation strategy through trial and error interaction, without relying on predefined maps. This paper summarizes the maze solving methods based on RL. The research methods mainly include seven categories: Deep Q-Networks, Quantum Reinforcement Learning(QRL), Hierarchical RL, Curriculum Learning, Reward Shaping, Transfer Learning, and Multi-Agent RL. Each category has representative works, potential mechanisms and experimental results. Then the sample efficiency, scalability, observability requirements and applicability are compared to determine three urgent challenges: sample efficiency in large-scale maze, partial observability and processing of dynamic environment, and the gap between simulation and reality. The purpose of this survey is to provide researchers with a systematic understanding of the current landscape and determine the direction of future work.
Weijia Li· Theoretical and Natural Scie...· 0 citations
