Robotics
Learning-Based Control
ActivePublished: 9 September 2026Updated: 9 September 2026Published
Key
innovation
Deriving control strategies from data and machine learning (including neural networks and reinforcement learning) rather than solely from analytically derived mathematical models.
Category
Robotics
Abstraction level
Paradigm
Operation level
Robot controlTraining
Use cases
Dynamic legged-robot locomotionDexterous manipulationAdaptive control in changing conditionsAutonomous vehicle control
How it works
A learning model (e.g. a neural network or RL policy) is trained on simulated or real data to map state observations to control actions; the policy is optimized against a reward/objective function and refined iteratively, often with simulation-to-reality transfer (sim-to-real).
Problem solved
For many systems (e.g. dynamic locomotion, contact, dexterous manipulation) an accurate analytical model is hard to derive; learning-based control obtains effective control strategies directly from data.
Key mechanisms
Reinforcement learning (RL)
Neural networks as controllers
Data-driven system identification
Sim-to-real transfer
Strengths & limitations
Strengths
✓Handles behaviors that are hard to model
✓Adapts to changing, complex conditions
✓Improves with data and experience
Limitations
✗Requires large datasets / costly training
✗Difficulty providing stability and safety guarantees
✗Risk of the sim-to-real gap and poor generalization