Braintech: Intelligent Vision

What the heck is Random Bin Picking?

It is a balmy afternoon on a rare sunny winter day in Vancouver and the hard working men and women of Braintech Canada are busy advancing the company's stack of intelligent vision technologies into the future. By way of introductions, Braintech Canada Inc. based in North Vancouver, British Columbia is the research & development member of the Braintech family and home to a majority of our world class team of vision and robotics research scientists, software engineers and quality assurance staff. Over the past 10 years, this diverse and highly talented group of people has been responsible for the development of many industry-first vision technologies and products including the award winning eVisionFactory solution studio for vision guided robotics and a host of other pioneering technologies such as the patented Single Camera 3D algorithms for object location.

Given the recent press release on the Random Bin Picking (RBP) technology I thought it may be appropriate to expand a little on the current developments in this area in my first blog entry. The RBP project was launched in late 2006 under a teaming arrangement between Braintech, ABB, Toyota, University of British Columbia and PRECARN (Canadian federal funding agency) with Braintech as the lead participant and PRECARN providing partial R&D funding. The development has to date produced a suite of patent pending robot vision technologies and techniques --collectively called RBP-- that plug into the eVisionFactory platform and in effect enable a robot to identify, locate and safely pick out individual parts from a randomly jumbled bin of objects. Essentially Braintech has been leading the development of the project and holds multiple patents pending on the technology.

At this point you are probably wondering just what the heck random bin picking is. To visualize a typical RBP application, think of a heap of engine pistons or a pile of cell phone covers thrown randomly into a box after being initially manufactured—by the way this is how most widgets get from point A to point B in a typical production process. Up until now, every time that cellphone cover or piston or other widget needed to be fed into a machine to be further processed (for example to be bored, painted or polished), this had to be done manually since a robot could not (even with existing vision technologies) see and find each piece in the jumbled heap and pick it out without colliding with the other pieces or the bin itself.

With RBP software, we essentially give the robot advanced 'eyes' to not only see individual parts but also to rank them and tell the robot which is the best one to pick: for example the cell phone cover that is highest in the pile or the one that is least obstructed by other cell phone covers is the best to pick. It may sound hard to believe, but this is in fact a very tough thing to get a computer to do and it's very exciting that we are now able to do this using the RBP algorithms. The key differentiator between RBP and existing bin picking technologies is that RBP was built from the ground up to deal with random bins of parts whereas existing solutions are typically adaptations of legacy vision solutions and even though can sometimes solve isolated random bin applications, they don't apply across the board or only work in structured environments.

Given the number of places where this technology can be used to save costs (think of how many widgets move around in factories in jumbled bins) and given our lead in this area, we are very focused on ensuring the successful roll-out of this technology starting with applications at select customers including Toyota Motor Manufacturing. We are also greatly thankful to Toyota and others who have played a pioneering role with us in this project right from the beginning by providing us with example applications, sample parts and statement of their needs. We look forward to working closely with these leaders to add the necessary enhancements, features and capabilities to meet their stringent plant floor requirements.

Best Regards,

Babak Habibi
CTO