An NPID Controller with Integrated Neural Network Could Help Bring BB-8-Style "Ballbots" to Reality
A tall robot balanced on a single ball may seem like a bad idea, but with this clever controller it's the image of stability.
Researchers from the Hanoi University of Industry, Queen’s University Belfast, the National Taiwan University of Science and Technology, the Thapar Institute of Engineering and Technology, and the Shibaura Institute of Technology have developed a robust, adaptive controller designed to bring BB-8-style robots to reality — or, as the team would have it, "ballbots."
"Ballbots with this advanced controller can be used as assistive robots for tasks requiring high mobility and precision," project lead Van-Truong Nguyen claims of the unusual form-factor. "For instance, they can assist individuals with mobility challenges in navigating complex environments. In addition, they can be used as service robots in dynamic settings such as restaurants, hospitals, or airports, offering smooth navigation. The robust self-balancing capabilities can be applied to delivery robots that need to operate efficiently despite unpredictable forces like wind or uneven terrain."
Like Star Wars' BB-8, a "ballbot" is based around a large central ball that contacts the ground — replacing the multiple wheels of a traditional robot design with a single ball, offering omnidirectional movement and a zero-point turning circle. The ball, in turn, is rotated using a series of motors controlled by the robot body balanced atop the ball — like BB-8's head, though considerable larger in scale.
Having such a tall design and therefore high center of gravity, though, makes these ballbots a challenge to control. That's where the team's controller design comes in: an adaptive non-linear proportional–integral–derivative (NPID) design with an integrated radial basis function neural network (RBFNN). This, the researchers claim, delivers improved stability, a reduction in "chattering" as the motors briefly fire to maintain balance, and high robustness gainst external disturbances — while requiring only light computational resources.
In testing, the team found its design outperforming both traditional PID and NPID controller types — including the ability to adapt to varying surface types using a self-learning and self-adjusting system not present in rival designs. The reduction in chattering and other unnecessary movements also delivered energy efficiency gains, the researchers claim — with Nguyen suggesting that "overall, industries such as logistics, healthcare, and retail could benefit from robots equipped with our technology, improving efficiency and service quality while reducing human workload."
The team's work has been published in the journal Engineering Science and Technology under open-access terms.
Main article image courtesy of the Shibaura Institute of Technology.