partial-observability

Online Replanning in Belief Space for Partially Observable Task and Motion Problems

SymDer: Symbolic Derivative Approach to Discovering Sparse Interpretable Dynamics from Partial Observations

Solving pursuit-evasion problems on graphs using Reinfocement Learning and GNNs

Official PyTorch implementation of POEM (Partial Observation Experts Modelling) as introduced in the paper Contrastive Meta-Learning for Partially Observable Few-Shot Learning

MATLAB code for the paper "Assimilative Causal Inference".

UWM-JEPA: a JEPA world model with a density-matrix latent and learned unitary predictor for imagining hidden continuations under partial observability.

Code for a multi-agent particle environment used in the paper "Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments"

Partially Observable Monte Carlo Planning algorithm (POMCP)

A high-performance, deterministic multi-agent simulation and benchmarking environment built from the ground up in C# and .NET 10 for faster research iteration.

Comparing a full-map A* oracle with partially observable knowledge-based and genetic-policy agents in Wumpus World.

Hybrid planning agent for decision-making under partial observability (PG-WMA)

Decentralized traffic signal control under camera-limited observability (IEEE Access submission)

Various projects related to contingent planning under partial observability

End-to-end empirical implementation of model-based off-policy evaluation and pessimistic policy selection for confounded POMDPs (Hong, Qi & Xu, ICML 2024)

Minimal control-theoretic toy models for studying stability and failure modes under partial observability.

Blog post on safety constraint learning

Defender-side cyber incident simulator. An abstract state machine, not real infrastructure, and reinforcement-learning policies, not LLM agents. Host status is hidden behind noisy alerts, the attacker improvises, and a web console lets you play the same scenario by hand and compare your score to a trained agent's.

Priced information-gathering and abstention evaluation in a synthetic partially observed city.

The AIS planner source code + benchmarks

意思を持つテトロミノたち — 各落下ピースが盤面スナップショット画像だけを見て収まり先を決めるマルチエージェント・テトリス PoC。ローカル VLA 対応、自律/A2A/独裁の協調モード比較、マシン性能ベンチマーク付き。