deep-reinforcement-learning

Berkeley's Deep RL class is now fully available online on YouTube for free, as announced by Sergey Levine.

The lectures for the Deep Reinforcement Learning course CS185/285 from UC Berkeley are now available online for public viewing.

This paper proposes a proxemics-based reward formulation for deep reinforcement learning social navigation, modeling human personal space as Gaussian-mixture fields to improve social compliance while maintaining navigation efficiency.

This paper presents a deep reinforcement learning approach for solving vehicle routing problems, demonstrated through three industrial truck planning case studies. The proposed method achieves over 10% cost reduction compared to baseline results and discusses generalization to more VRP variants.

This paper proposes PLAN, a lightweight parallel liquid-inspired approximation network for efficient representation learning in flexible job shop scheduling, achieving better makespan and lower inference latency with fewer parameters than state-of-the-art baselines.

This paper reports that deep reinforcement learning agents using frozen, randomly initialized CNN feature extractors spontaneously develop extremely sparse fully-connected representations, compressing task-relevant information through very few neurons without any sparsity-inducing objective.

This survey provides a unified view of progress reward modeling for robotic learning, organizing the field into three steps: interface, methods, and data/benchmarks.

This paper explores Relative Positional Encoding (RPE) as an additive bias in Transformer architectures to solve the Team Orienteering Problem, demonstrating consistent improvements in collected rewards and optimality gaps over vanilla Transformer architectures.

This paper systematically explores four deep reinforcement learning solutions (DQN, REINFORCE, PPO, and MuZero) for the asymmetric Nepali board game Baghchal, finding that MuZero achieves the best win rates due to model-based planning via Monte Carlo Tree Search.

Proposes a two-timescale multi-layer deep reinforcement learning framework with latent action space for joint service placement, computational delegation, and power control in hierarchical edge-cloud computing, achieving up to 20.8% latency reduction and 13% resource utilization improvement.

This paper analyzes the evaluation and design paradigms in deep reinforcement learning, revealing that performance rankings are not monotonic across data regimes and that common low-data regime benchmarks may lead to incorrect conclusions.

This paper proposes a deep reinforcement learning framework (MORP-DRL) for multi-objective reliability-based portfolio optimization, jointly optimizing expected return and downside risk using CVaR and EVaR under practical constraints, and demonstrates performance on global equity indices across different market regimes.

This paper proposes a Proximal Policy Optimization (PPO)-based deep reinforcement learning framework for dynamic battery charging of autonomous mobile robots in warehouses, achieving up to 6% higher order-completion rates over baseline methods.

This paper analyzes evaluation and design paradigms in deep reinforcement learning, demonstrating that canonical paradigms can lead to incorrect conclusions and providing insights into scaling, capacity, and complexity.

This work-in-progress paper proposes embedding a differentiable physics model into the PPO actor loss function to penalize anticipated safety violations in reinforcement learning, evaluated on a simulated 1-DoF helicopter system. The physics-informed soft regularizations reduce constraint violations while maintaining reliable target tracking.

This paper presents PASE, a neuro-symbolic framework that uses LLMs to generate structured recovery plans for cloud systems and verifies them via a neural-symbolic world model, achieving over 40% reduction in recovery time.

A three-phase deep reinforcement learning system for personalized portfolio management that addresses ticker lock-in, monolithic objectives, and static user models, using a cross-asset encoder pretrained with self-supervised learning and the Chronos time series foundation model, fine-tuned with Mixture of Experts and PPO, and personalized via LoRA.

This paper proposes a novel offline reinforcement learning framework for active flow control that uses a sensor position-conditioned architecture with Point Attention layers to handle varying sensor configurations, enabling data-driven policy extraction without costly online interactions.

Introduces A3M, a framework combining adaptive deep reinforcement learning, adversarial reasoning, and multi-objective reward design for strategic bidding in repeated auctions, achieving 30-40% regret reduction.

This paper introduces CARLOS, a deep reinforcement learning algorithm that learns continuous-time optimal stopping rules for American-style options using an aggregate deep neural network, effectively closing the Bermudan-American value gap with high computational efficiency.