A Markov decision process is the standard maths framework for reinforcement learning: states, actions, rewards, and the chances of moving from one state to the next.
In plain words
An MDP describes a world an agent acts in. It lists the possible states, the actions available, the reward for each move, and the probability that an action in one state leads to each next state. The Markov part means the next state depends only on the current state and action, not on the whole history. The agent's aim is to collect as much total reward as possible.
A quick example
Imagine a robot vacuum. Its state is where it is and how much battery is left; its actions are move or recharge; it earns reward for cleaned floor. Framing this as an MDP lets a reinforcement learning method learn a good cleaning policy.
Where you'll meet it
- Formulating reinforcement learning problems.
- Planning and decision-making under uncertainty.
Related
Sources
- DocsMachine Learning Glossary, Google for Developers · read 29 Sept 2026
- DocsPart 1: Key Concepts in RL, OpenAI Spinning Up · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.