SFIT competed in the inaugral Amazon Picking Challenge 2015 held at ICRA
2015 in Seattle, USA from 26-29 May, 2015. The system aimed at automating
the process of picking up objects from the stocking shelf in huge warehouses
like the ones used by Amazon.
APC for team SFIT was a platform to compete at an international
level and work on a real world problem faced in the warehousing industry.
Compared to other teams at APC, team SFIT lacked funds and comprised only
of 5 members (inclusive of faculty advisor). This forced them to be extremely
creative and adopt the frugal innovation strategy. Our solution won the
travel grant and was applauded for its ingenuity at the challenge.
This competition had plenty of challenges thrown at us. The multiple
challenges included object detection, object recognition, motion planning,
pose and orientation adjustments, gripping, etc. While most of the fellow
competitiors used an industrial platform like the PR2 Robot, ABB Robot
or the famous research platform Baxter robot, we came up with an absolutely
ingenious solution. Instead of using a single huge robotic arm, we decided
on using 12 small individual robots which would fetch items from each shelf
and then with a smart combination of conveyour systems, the objects were
moved to the order bin. Checkout the innovative approach taken by team
SFIT.
The challenge required each robot to be able to pick all the 25 items without damaging them. The gripper design became essential in this aspect. Additionally, the small size of the bin restricted the size of grippers. Overall I was able to design and fabricate the robots which could take care of all these aspects. Each robot was equipped with a camera, a pair of grippers, motors and the circuit to drive the robot. The robot was externally powered. The processing was offloaded to individual raspberry pi's.
A quick look at the proceedings of the challenge.