Remember meForgot password?
    Log in with Twitter

article imageCollision detector to make robots better

By Tim Sandle     Nov 29, 2017 in Science
The application of speedy collision detectors could make robots better human assistants, according from a study conducted at the University of California - San Diego.
The new research into robotics has concluded that a faster type of collision detection algorithm would have the capability of allowing robots to work more fluidly. Applications for modified robotic assistants include the operating room or care homes, designed to add support for those requiring assisted living.
The University of California algorithm is called 'Fastron' (a combination of the words Fast and Perceptron), and in tests it has been shown to run up to eight times faster than comparable collision detection algorithms. The code takes advantages of machine learning to guide robots into avoiding moving objects. The algorithm also enables robots to weave through complex, rapidly changing environments, sensing their surroundings in real time.
The algorithm was presented at the world's first annual "Conference on Robot Learning", which took place during November 2017 at Google headquarters in Mountain View, California, U.S. At the conference the surgical application was discussed. Here it was explained that the new algorithm is suitable for use with the da Vinci Surgical System. This consists of a robotic arm that can autonomously perform supporting tasks for surgeons, such as suction, irrigation or pulling tissue back.
For the research, the technologists worked at the Advanced Robotics and Controls Lab at the university, and they adopted a so-termed minimalistic approach to collision detection. The output - the Fastron algorithm - makes use of machine learning strategies which were initially developed to classify objects. From this the robot can lean to classify collisions versus non-collisions in dynamic environments.
A supporting white paper has been produced, titled "Fastron: An Online Learning-Based Model and Active Learning Strategy for Proxy Collision Detection."
More about Robots, Robotics, Machines, assistants
More news from