Simulation and AI help Leeds robot conquer clutter

Engineers at the University of Leeds have combined automated planning with machine learning to help a robotic arm deal with cluttered environments.

Object manipulation and grasping are notoriously difficult for robots, especially when the target is surrounded by other objects. For humans, planning a route for your hand through a cluttered table in order to pick up an apple is second nature. For a robot, the computation needed may be so extreme that it takes minutes to figure out its strategy, and even then it will often fail in the task.

US robotics teams gear up for DARPA mine challenge

Robotic hand gets to grips with assembly line jobs

By combining two different approaches to this problem, the Leeds team taught a robotic arm to deal with these situations autonomously and efficiently. Automated planning uses computer vision and software to simulate the sequence of moves that might be used to achieve the goal. These simulations can be useful, but don’t replicate the complexity of performing the operations in the real world. To counter this, the researchers also used machine learning to train the robot in around 10,000 trial and error situations to refine its technique.

“Artificial intelligence is good at enabling robots to reason – for example, we have seen robots involved in games of chess with grandmasters,” said Dr Matteo Leonetti, from the university’s School of Computing.

“But robots aren’t very good at what humans do very well: being highly mobile and dexterous. Those physical skills have been hardwired into the human brain, the result of evolution and the way we practise and practise and practise. And that is an idea that we are applying to the next generation of robots.”

(Credit: University of Leeds)

In research that will be presented at the International Conference on Intelligent Robotics and Systems in Macau, China, the robot performed a serious of dexterity tasks on a cluttered table, moving various objects to target areas.

“Our work is significant because it combines planning with reinforcement learning,” said Wissam Bejjani, the Leeds PhD student who authored the paper.

“A lot of research to try and develop this technology focuses on just one of those approaches. Our approach has been validated by results we have seen in the university’s robotics lab. With one problem, where the robot had to move a large apple, it first went to the left side of the apple to move away the clutter, before manipulating the apple. It did this without the clutter falling outside the boundary of the shelf.”