March 04, 2026
Good motion planning is what makes a robot feel reliable. When the environment changes, when a depth sensor sees noise, or when the robot works in a tight space, the planning stack decides whether the system reacts with confidence or with guesswork. At PickNik, we’ve supported teams building robots for space, labs, and factory floors. Along the way we’ve used many 3D mapping tools, including our Octomap-based approach in MoveIt Pro and NVIDIA’s NVBlox + CuMotion stack.
This post explains how the two systems differ and why many teams still pick Octomap for real deployments.
Mapping the World: Octomap and NVBlox
Octomap builds a probabilistic 3D occupancy map from depth data. It handles noisy sensors well and \offers easy API libraries to manage the map’s resolution, clearing, and update rules. It plugs directly into the MoveIt Pro planning scene, so the integrated planners can work from it’s consistent model of the world.
NVBlox uses a TSDF to create a mesh and is optimized for GPU speed. It can run very fast, but it needs careful tuning when the sensor data is messy. Without that tuning, small reflections or stray points degrade the map quality meaning a loss in planning time and correctness.
Mapping the World: Octomap and NVBlox
Octomap builds a probabilistic 3D occupancy map from depth data. It handles noisy sensors well and \offers easy API libraries to manage the map’s resolution, clearing, and update rules. It plugs directly into the MoveIt Pro planning scene, so the integrated planners can work from it’s consistent model of the world.
NVBlox uses a TSDF to create a mesh and is optimized for GPU speed. It can run very fast, but it needs careful tuning when the sensor data is messy. Without that tuning, small reflections or stray points degrade the map quality meaning a loss in planning time and correctness.
How the Planners Behave
MoveIt Pro uses our custom planners based on RRT that respect constraints and understand the arm’s workspace. With Octomap in the loop, the system is optimized to pick short, safe motions and avoids wide detours, while not sacrificing the planning speed. This matters when the robot operates near fixtures, tools, or people.
CuMotion is NVIDIA’s GPU planner, allowing for planning to be accelerated when using NVIDIA compute devices. Many users tell us its motions can feel abrupt or inefficient. In one comparison we saw joint motions that were larger than needed and didn’t match the workspace well, especially when paired with NVBlox.
What We See in the Field
A recent customer ran both stacks on a UR robot in a Polyscope X cell. Their depth camera produced reflections that confused the map. After they tuned Octomap’s filters, the planning scene became clean, and the first Octomap plan executed safely on hardware.
Their takeaway was simple: “This planning is much better than NVIDIA.”
We’ve seen the same pattern during on-site support. NVBlox can produce detours or unstable behavior when the space is tight, while Octomap tends to remain steady once tuned. Thanks to Shaur for helping that customer land on a working Octomap setup when it mattered.
Why Teams Stick With MoveIt Pro
MoveIt Pro is more than a planner. It’s a full manipulation stack with real-time updates, behavior trees, sensor-driven autonomy, and tools for deploying production systems. Our Octomap integration has gone through years of testing in many environments. It’s predictable, transparent to tune, and works well when the robot needs to react to new obstacles on the fly.
If you want your robot to produce smooth, consistent motions without surprises, the Octomap-based stack in MoveIt Pro is a strong choice. And if you’re building a system that has to work on the first try, this matters.
MoveIt Pro is PickNik Robotics’ commercial robotics software platform that extends the capabilities of the open-source MoveIt project. Designed for unstructured environments, MoveIt Pro provides powerful tools for motion planning, grasping, perception, and behavior sequencing. With pre-built robot behaviors and a modular runtime architecture, MoveIt Pro significantly reduces engineering effort and time-to-deployment. Its hardware-agnostic design supports a wide range of commercial and custom robotic arms, making it ideal for scalable automation across space, manufacturing, and logistics sectors.
