A recognition engine and application-building software for random or structured bin picking 

NebulaRecognition® + Planner

NebulaRecognition + Planner is our latest technology designed to help you easily set up a random or structured robotic bin picking application which utilizes an advanced object recognition algorithm paired with collision avoidance. 



NebulaRecognition is our intelligent and lightweight recognition engine that learns from point cloud samples and locates the model anywhere in the scene, regardless of its location (X, Y, Z) and orientation (Rx, Ry, Rz). The optimized algorithms of NebulaRecognition allow recognition to run fast — even on low power CPUs.  

Unlike competition, NebulaRecognition does not require CAD files to learn. Additionally, many competitive object recognition techniques involve segmentation or deep neural networks, which are computationally costly and require “hand crafting” thus making them an engineered and unscalable solution. 

NebulaRecognition can rapidly: 

  • Be taught from point cloud samples  
  • Learn from CAD files 
  • Locate multiple different taught objects placed and oriented randomly in a scene 
  • Find surfaces of any dimension, with both manual input and Automatic Learning 

To teach an object, simply trigger the sensor, click to teach, and erase the background.


The Terrace is the sensor behind NebulaRecognition + Planner’s high-definition point cloud generation. The Terrace is qualified for both robot-mounted and fixed sensor applications. 


Planner is our easy-to-use bin picking application software designed to optimize picking and avoid collisions in the application environment. It’s a 3D robotic cell simulator that communicates directly with your robot, guiding it to pick parts located by NebulaRecognition. 

The process: NebulaRecognition gives Planner location data about all the parts it has recognized, then Planner rapidly checks for collisions, solves your robot’s inverse kinematics, and chooses the most optimal part to pick, avoiding collisions and minimizing robot movement. 

Planner can: 

  • Detect collisions between the robot, tooling, and other user-defined objects in the scene 
  • Solve inverse kinematics for supported robot brands and models 
  • Automatically computes hand-eye coordinates via our own Toolfinder® 
  • Teach positions directly 
  • Teach picking locations of the object 
  • Track TCP position in real time during operation 
  • Display scene point clouds and measured data during a live session 
  • Add visual shapes into the virtual scene to mimic the real-life set up 
  • Support two-sensor setups 

and much more. 

One of Planner’s most helpful features is Sequencer, which describes your application’s sequence step-by-step. Sequencer’s node-based structure allows you to edit position, acquisition, object type branching, and tooling nodes directly, thus reducing the need for robot programming. In fact, most robot-mounted sensor applications can be run entirely with Sequencer and zero to little robot programming. 

Some of Sequencer’s functionality includes: 

  • Setting acquisition types 
  • Modelling robot location at taught positions  
  • Controlling robot speed to and from positions 
  • Predict recognition and picking decisions  
  • Providing real-time visualization of the sequence as your application is running


Sensor Specifications Sheet

NebulaRecognition + Planner is an easy-to-use, highly customizable, and intelligent bin picking solution. Request a live demo today! 

Download PDF

Copyright © Recognition Robotics Inc. | All Rights Reserved