Robot control
Central pattern generator
Definition
A central pattern generator is a neural circuit or mathematical oscillator network that produces coordinated rhythmic commands from compact control inputs. In robotics, CPG models are used to generate and modulate repeated motions such as walking, swimming, or wing beating.
Also known as: CPG, Central pattern generators
Updated
Generate a rhythm from a compact command
A central pattern generator produces structured, repeating output without needing a separate command for every point in the cycle. Several coupled oscillators can represent the phases of different legs or joints. A small set of inputs can then change frequency, amplitude, phase relationships, or gait.
Ijspeert's review describes biological CPGs as neural circuits that create coordinated, high-dimensional rhythmic signals from simpler inputs. It also reviews robotic implementations based on neural networks and coupled oscillators for articulated locomotion.
Rhythm and feedback work together
A robot can use a CPG as a reference generator while sensory feedback changes its phase or output. Foot contact, joint load, body tilt, or proprioceptive measurements can help synchronize the rhythm with the physical motion. This differs from replaying a fixed periodic trajectory, which does not adjust its timing when a foot touches down early.
A CPG is also not the same as a dynamic movement primitive. Both can generate motion through dynamical systems, but CPGs specifically emphasize sustained rhythmic coordination. A movement primitive may instead converge to a single goal or encode a non-periodic skill.
An oscillator does not ensure a stable gait
The generated pattern still has to respect joint limits, actuator bandwidth, contact forces, and balance. A set of synchronized oscillators can command a clean rhythm while the physical robot slips or falls. A whole-body controller or other feedback layer may still be needed to manage contacts and disturbances.
Oscillator parameters can also be difficult to tune across speeds, payloads, and terrain. Strong sensory coupling may improve adaptation but can disrupt the intended phase relationships or create unwanted oscillations. Evaluation should therefore separate the quality of the rhythmic reference from stability and task performance on the robot.
Sources
Related terms
Bipedal locomotion
Bipedal locomotion is movement using two legs, with body motion coordinated through changing contacts between the feet and the environment. It includes walking and running.
Dynamic movement primitive
A dynamic movement primitive is a parameterised dynamical system that represents a goal-directed or rhythmic movement using stable baseline dynamics plus a learned shaping term. Robots can fit the parameters from demonstrations and adapt the resulting motion to a new goal or duration.
Whole-body control
Whole-body control coordinates a robot’s joints and contacts to satisfy several motion and force objectives together. In humanoids, it commonly combines balance, foot motion, hand tasks, and posture subject to physical constraints.
Proprioception
Proprioception in robotics is sensing the robot's own motion, configuration, and internal physical state. Typical proprioceptive inputs include joint encoders, inertial measurements, and signals associated with actuator effort or contact.