This review discusses the application of optimality principles in sensorimotor control. The sensorimotor system is shaped by evolution, development, learning, and adaptation, which work on different time scales to improve performance. Many theories of motor function are based on the idea of optimal performance, using optimal control theory to predict behavior. These models have explained a wide range of empirical phenomena better than other models. Traditional models focused on average trajectories, ignoring feedback, but recent work has redefined optimality in terms of feedback control laws and online behavior mechanisms. This has led to a unified theoretical framework for motor function, emphasizing the relationship between high-level goals and real-time control strategies. Optimality principles are central to many scientific theories, as they allow for the transformation of a simple performance criterion into detailed predictions. Optimal control models of biological movement have successfully explained behavioral observations at multiple levels, including limb trajectories, joint torques, and muscle activations. These models are both theoretically and practically advantageous, as they are well-justified a priori due to the evolutionary nature of the sensorimotor system. They also provide autonomy and generality, as they only require a performance criterion and automatically fill in movement details. The review discusses different types of optimal control models, including open-loop and closed-loop models. Open-loop models predict average behavior by optimizing cost functions, but have limitations in modeling trial-to-trial variability and online feedback. Closed-loop models, on the other hand, consider motor and sensory noise, and predict both average behavior and sensorimotor contingencies. These models have unified various concepts and observations into a cohesive framework, and may allow for future extensions such as hierarchical control and automated task goal inference. The review also explores different cost functions used in optimal control models, including energy, smoothness, and accuracy. Energy minimization is common in biomechanics, while smoothness optimization is successful in predicting average trajectories. Accuracy optimization minimizes variance in final positions. Multi-attribute cost functions combine different aspects of performance, such as accuracy and energy, and can be used to predict muscle directional tuning. The review concludes that optimal feedback control provides a natural framework for studying sensorimotor function, and that hierarchical models of sensorimotor control may be useful for understanding complex behaviors. These models incorporate feedback loops at different levels, and may be applicable to real-time control of robotic prostheses and electrical stimulators. The review also highlights the importance of considering low-level feedback loops in understanding motor behavior.This review discusses the application of optimality principles in sensorimotor control. The sensorimotor system is shaped by evolution, development, learning, and adaptation, which work on different time scales to improve performance. Many theories of motor function are based on the idea of optimal performance, using optimal control theory to predict behavior. These models have explained a wide range of empirical phenomena better than other models. Traditional models focused on average trajectories, ignoring feedback, but recent work has redefined optimality in terms of feedback control laws and online behavior mechanisms. This has led to a unified theoretical framework for motor function, emphasizing the relationship between high-level goals and real-time control strategies. Optimality principles are central to many scientific theories, as they allow for the transformation of a simple performance criterion into detailed predictions. Optimal control models of biological movement have successfully explained behavioral observations at multiple levels, including limb trajectories, joint torques, and muscle activations. These models are both theoretically and practically advantageous, as they are well-justified a priori due to the evolutionary nature of the sensorimotor system. They also provide autonomy and generality, as they only require a performance criterion and automatically fill in movement details. The review discusses different types of optimal control models, including open-loop and closed-loop models. Open-loop models predict average behavior by optimizing cost functions, but have limitations in modeling trial-to-trial variability and online feedback. Closed-loop models, on the other hand, consider motor and sensory noise, and predict both average behavior and sensorimotor contingencies. These models have unified various concepts and observations into a cohesive framework, and may allow for future extensions such as hierarchical control and automated task goal inference. The review also explores different cost functions used in optimal control models, including energy, smoothness, and accuracy. Energy minimization is common in biomechanics, while smoothness optimization is successful in predicting average trajectories. Accuracy optimization minimizes variance in final positions. Multi-attribute cost functions combine different aspects of performance, such as accuracy and energy, and can be used to predict muscle directional tuning. The review concludes that optimal feedback control provides a natural framework for studying sensorimotor function, and that hierarchical models of sensorimotor control may be useful for understanding complex behaviors. These models incorporate feedback loops at different levels, and may be applicable to real-time control of robotic prostheses and electrical stimulators. The review also highlights the importance of considering low-level feedback loops in understanding motor behavior.