Reinforcement Learning
Il Reinforcement Learning è un approccio di apprendimento computazionale orientato agli obiettivi in cui un agente impara a svolgere un’attività interagendo con un ambiente dinamico sconosciuto. Durante l'addestramento, l'algoritmo di apprendimento aggiorna i parametri della politica dell'agente. L'obiettivo dell'algoritmo di apprendimento è individuare una politica ottimale che massimizzi la ricompensa cumulativa attesa e scontata a lungo termine ricevuto durante l’attività.
Questo approccio di apprendimento consente all'agente di prendere una serie di decisioni volte a massimizzare la ricompensa complessiva per un'attività, senza l'intervento umano e senza essere stato esplicitamente programmato per raggiungere un obiettivo. È possibile creare e addestrare agenti di Reinforcement Learning utilizzando il software Reinforcement Learning Toolbox™.
Per ulteriori informazioni, vedere What Is Reinforcement Learning? (Reinforcement Learning Toolbox).
Argomenti
- What Is Reinforcement Learning? (Reinforcement Learning Toolbox)
Reinforcement learning is a goal-directed computational approach where a computer learns to perform a task by interacting with an uncertain dynamic environment.
- Reinforcement Learning Workflow (Reinforcement Learning Toolbox)
Typical workflow you use to apply reinforcement learning to a problem.
- Reinforcement Learning Environments (Reinforcement Learning Toolbox)
Model environment dynamics using a MATLAB® object that generates rewards and observations in response to agents actions.
- Reinforcement Learning for Control Systems Applications (Reinforcement Learning Toolbox)
You can train a reinforcement learning agent to control a plant.
- Train Reinforcement Learning Agent in MDP Environment (Reinforcement Learning Toolbox)
Train a reinforcement learning agent in a generic Markov decision process environment.
- Train Reinforcement Learning Agent in Basic Grid World (Reinforcement Learning Toolbox)
Train Q-learning and SARSA agents to solve a grid world in MATLAB.
- Design and Train Agent Using Reinforcement Learning Designer (Reinforcement Learning Toolbox)
Design and train a DQN agent for a cart-pole system using the Reinforcement Learning Designer app.
- Create DQN Agent Using Deep Network Designer and Train Using Image Observations (Reinforcement Learning Toolbox)
Create a reinforcement learning agent using the Deep Network Designer app from the Deep Learning Toolbox™.
- Train DDPG Agent with Custom Networks Using Image Observation (Reinforcement Learning Toolbox)
Train a DDPG agent with custom networks using an image-based observation signal.
- Control Water Level in a Tank Using a DDPG Agent (Reinforcement Learning Toolbox)
Train a controller using reinforcement learning with a plant modeled in Simulink® as the training environment.
Esempi in primo piano
Train DQN Agent for Lane Keeping Assist Using Parallel Computing
Train a DQN agent for an automated driving application using parallel computing.
Deep Reinforcement Learning for Optimal Trade Execution
Use the Reinforcement Learning Toolbox™ and Deep Learning Toolbox™ to design agents for optimal trade execution.
Train Biped Robot to Walk Using Reinforcement Learning Agents
Compare DDPG and TD3 agent for the control a biped walking robot modeled in Simscape™ Multibody™.
Train DDPG Agent for Adaptive Cruise Control
Train a DDPG agent for an adaptive cruise control application.
Train Hybrid SAC Agent for Path-Following Control
Train a hybrid SAC agent for lane following control.
Automatic Parking Valet with Unreal Engine Simulation
Use a TD3 agent with an MPC controller to perform a parking maneuver.
Train Humanoid Walker
Model a humanoid robot using Simscape Multibody™ and train it using either a genetic algorithm (which requires a Global Optimization Toolbox license) or reinforcement learning (which requires Deep Learning Toolbox™ and Reinforcement Learning Toolbox™ licenses).
Train Policy Deployed on Raspberry Pi
Use trainFromData in a MATLAB learning loop to train a policy
deployed on a Raspberry Pi board.
