scooter cobot for guiding horizontal movements in position and orientation
 
 

Excellent Students wanted for this project in collaboration with Northwestern University. Get a NUS degree and US experience in a leading Robotics lab!
 
 

Learning Cobots

Cobots are mechatronic devices invented by E. Colgate and M. Peshkin, Northwestern University realizing mechanically software defined surfaces. Cobots create virtual surfaces that an operator can employ in performing a material-handling task. These surfaces provide smooth guidance for the motion of a payload, yet let the operator free to move the payload at will when it is not in contact with a virtual surface. Since virtual surfaces are defined in software, they can be applied liberally wherever needed in a workspace, and can be adjusted on the fly for mixed-product lines. More information and papers about cobots can be found here.

The goal of this project is to develop a simple method to program surfaces resulting in an optimal collaboration of the cobot and the human operator. Using a classical optimization framework to compute these surfaces may be not very useful, as the forces exerted by the operator during movement and his/her comfort can hardly be considered. This project's idea is to put the operator in the loop. We will provide him/her ways to modify and improve the cobot command trial after trial. The surfaces will be coded as B-splines, i.e. be defined as a linear sum of "attraction points". Two ways will be investigated to design and modify the surface: i) by recording movements performed by th e operator ("teach pendant mode") and ii) by using a computer mouse to place/move attraction points on a monitor on which the path and the environment are represented ("CAD mode"). Possible combination of these two modes may also be examined.

As for conventional robots, the teach-pendant probably is the simplest way to define a surface. During the initial movement, the cobot will be in free mode. The path driven will define the guiding surface that will be used from the second trial on. The operator will be able to bend this surface by pushing/pulling it, i.e. by applying a force on the surface. The deformation magnitude can be tuned using a parameter, and the operator will decrease this parameter to "stiffen" the surface when he/she feels that it is becoming optimal.

The benchmark task will consist of putting a pin fixed on the "scooter" cobot (above figure) into a fixed hole. The developments may be useful for spatial cobots also, e.g. for executing precision tasks using Carl Moore's arm cobot. The cobot and corresponding forces will first be emulated using a 3 DOF haptic interface and a computer monitor for visual display in Singapore. Then experiments will be performed at Northwestern University to test and extend the strategies.

back to Etienne BURDET homepage

January 2000, by Etienne Burdet