PlaCo (Rhoban planning and control library)

Rhoban's open-source planning and control library, which expresses IK and whole-body control as tasks solved by a QP.

PlaCo (Planning & Control) is an open-source library from the Rhoban team at the LaBRI lab of the University of Bordeaux, France. It's MIT-licensed, implemented in C++ with Python bindings, and on Linux and macOS installs with a plain pip install placo. Built on the rigid-body dynamics library Pinocchio, it lets you declare tasks and constraints — reach a pose with the end effector, keep the center of mass inside the support polygon, respect joint limits, avoid self-collision — then linearizes them around the current configuration using Jacobians, assembles a quadratic program (QP), and solves for the joint increment. Each task can be a hard constraint that must hold or a weighted soft objective. Beyond inverse kinematics, it offers an inverse-dynamics solver that handles contacts and torque limits, plus a humanoid walking pattern generator based on the linear inverted pendulum; the official examples include Rhoban's own Sigmaban humanoid. Newcomers usually run into it through Open Duck Mini, whose reference-motion generator uses PlaCo to produce the walking motions an RL policy learns to imitate.

ExampleUse the Open Duck reference-motion generator (built on PlaCo) to batch-generate walking reference motions across a range of forward, sideways, and turning speeds, then train a policy in MuJoCo Playground to imitate them.

Also called

placo

Related

Pinocchio · Quadratic Programming · Whole-Body Inverse Kinematics · Whole-Body Control · Pink · Open Duck Mini

Sources

Rhoban/placo (GitHub)
PlaCo: a QP-based robot planning and control framework (arXiv)
apirrone/Open_Duck_reference_motion_generator (GitHub)

As of

2026-10