ManipulaPy Documentation#

Move from equations to motion.#

Python robotics, from model to motion

Build, differentiate, and accelerate robot kinematics, dynamics, planning, perception, and control in Python.

Start building

python -m pip install manipulapy

Six-axis industrial robot arm in a research lab

Choose your path#

Learn robotics

Work through the mathematics

Follow executable notebooks from rigid transforms through differentiable dynamics.

Load a robot

Start from a URDF

Build the serial model, dynamics, limits, and frames from your robot description.

Plan motion

Turn goals into trajectories

Generate joint and Cartesian paths, then connect planning to control and simulation.

Accelerate and differentiate

Choose the right array backend

Run the same core mathematics on NumPy, CuPy, PyTorch, or JAX.

See the math move#

Robot arm tracing its reachable workspace — Explore reachable workspaces

Joint position, velocity, and acceleration changing along a trajectory — Inspect a smooth joint trajectory

UR5 robot executing a pick motion — Execute planned motion in simulation

Compute where your work belongs.#

One API, four array libraries

NumPy

The lightweight default for local work, teaching, and dependable CPU execution.

CuPy

Move compatible array work to NVIDIA GPUs without changing the public mathematics API.

PyTorch

Connect robot models to training loops and preserve gradients through supported core math.

JAX

Differentiate and compile supported kinematics, dynamics, singularity, and utility operations.

Read the backend guide Open the backend API

Go straight to the reference#