Benchmark RL environment for infrastructure maintenance planning

Quantile Regression DQN implementation for bridge fleet maintenance optimization using Markov Decision Process. Migrated from C51 distributional RL (v0.8) with 200 quantiles and Huber loss. Features: Dueling architecture, Noisy Networks, PER, N-step learning. All 6 maintenance actions show positive returns with 68-78% VaR improvement.

Deep Q-Network implementation for optimal bridge maintenance planning using Markov Decision Process formulation with vectorized parallel training. Based on Phase 3 (Vectorized DQN) from dql-maintenance-faster project.

A deep reinforcement learning system for optimizing bridge maintenance decisions across municipal infrastructure fleets, implementing cross-subsidy budget sharing and cooperative multi-agent learning.

Deep Q-Network (DQN) implementation for optimal maintenance planning of 100-bridge fleet infrastructure using advanced reinforcement learning techniques and vectorized parallel training.

This tool applies self-improving (Agentic) clustering to bridge maintenance data in Open data at some Prefecture, Japan, to automatically identify bridge groups with high maintenance priority.

This project applies self-improving (Agentic) clustering with Bayesian Optimization to bridge maintenance data in some Prefecture, Japan, to automatically identify bridge groups with high maintenance priority.

This system analyzes bridge repair method recommendation reports generated by AI agents and visualizes the decision-making pathway from damage → deterioration factors → repair methods as a Decision Tree. It aims to "make the thought process visible."

An AI-powered bridge health classification system that automatically categorizes bridge inspection reports into health levels using machine learning. The system leverages Explainable Boosting Machine (EBM) to achieve high accuracy while maintaining interpretability.

gym-based RL environment for infrastructure maintenance planning

C51 Distributional DQN (v0.8) for bridge fleet maintenance optimization. Implements categorical return distributions (Bellemare et al., PMLR 2017) with 300x speedup via vectorized projection. Combines Noisy Networks, Dueling DQN, Double DQN, PER, and n-step learning. Validated on 200-bridge fleet: +3,173 reward in 83 min (25k episodes).

This project contains an implementation of a decision tree model for maintenance and deterioration