"Teachingless" robotics projects...

In the case of conventional industrial robot systems, operators must write robot-language programs for each robot task by using hand-held "teach pendant" or off-line robot programming software. This process is long, tedious and requires technical experts in robot programming. Moreover, as each new robot motion needs a human action, the use of robots is restricted to repetitive tasks which need to be executed thousands of times. No variation around the recorded motion is allowed.

Performing numerous different tasks with the same flexible robot is becoming possible with "teachingless" solutions; opening new fields of use where programming the robot was economically not viable.

“Coupling hardware sensors to our automatic path planner software creates the bridge between virtual and real environments; robot motion is computed in the virtual representation of the real world and executed in this later without programming” says Laurent Maniscalco, Kineo CAM’s CEO; “as of today five of our customers have already or are currently developing teachingless robotic systems in the following fields: programming-less structural steel welding in North America, electronic assembly in Japan, automated part sorting in Germany or aircraft structure assembly control in France”.

Kineo CAM proposes a set of libraries to make industrial robots more autonomous, flexible and easy-to-use, extending the use of industrial robots to non-robotic experts for flexible tasks.

The software component named KineoWorks™ can be used offline within a dedicated system that will automatically compute the robot program before sending it to the shop floor. The automation of the programming permits quick reconfigurations of the robotic cell in a flexible manufacturing paradigm.

It can also be used online. Embedded on the system controlling the robot and coupled with a perception system (LIDAR, Structured light 3D scanner, computer vision...) and/or a Digital Mockup, it can automatically plan collision-free motion, control its execution, react to unexpected obstacles (moving object, worker, obstructed obstacle) and re-plan a detour on-the-fly. This is what we call reactive path planning:

A preliminary scanning of the environment allows planning an initial trajectory. Once the robot starts to execute the motion, KineoWorks™ continues to analyze the captured data as it comes in order to anticipate potential collisions. A continuous process tries to keep the obstacles away by deforming the trajectory. When the deformation fails, a new path planning is launched on-the-fly in parallel and without interrupting the ongoing motion.

Thanks to reactive and fast path planning, the robot can autonomously move in an unknown environment. In particular a project funded by the French government uses KineoWorks™ embedded into a UGV to explore unknown or “not well known” installations, such as a plant damaged by a natural disaster like an earthquake.

Automatic control of mechanical assys

Noomeo and Kineo CAM are collaborating to develop a solution for automatic control of the quality of mechanical assemblies.

Nowadays, most products that surround us are designed with computer-aided design (CAD) software. Resulting digital models (CAD models) are then used for the development of means of production that will enable the manufacture of parts.

One major difficulty is then to ensure that the result of production conforms to the original design. Non Destructive Testing (NDT) is a set of methods and techniques of perception and action implemented to verify such compliance. The origins of manufacturing errors are numerous: poor adjustment of machinery, tool wear ... but also, human operator errors when performing manual assembly, or when preparing machining or assembly performed by machines or robots.

In this context, Noomeo and Kineo CAM are collaborating to develop a solution called CAAMVis aiming at automating the quality control of mechanical assemblies by Artificial Vision. No learning phase is required other than loading the CAD model and identification of control points. Time savings are even greater when production lines must adapt to frequent configuration changes.

The assembly is inspected by the CAAMVis control system whose architecture consists of a robot arm equipped with a 2D/3D vision sensor, software for image processing and automatic comparison with stored digital models and software for automatic computation of 3D control sequences that are executed by the robot.

The main operations addressed by CAAMVis are:

• automatic selection of control points computed from the model and the current state of the assembly,
• motion planning of part holding and / or sensor holding manipulators allowing to reach previously computed control points,
• execution of generated robot trajectories,
• automatic comparison between the digital model and measurements made at each stage to check the compliance of the assembly.