Customized brains for robots
Customized brains for robots, is the new niche in the world for robotics. Though our existing robots do not lack speed for physical mobility, but their interactive abilities with humans need to touch higher notches. Sabrina Neuman is a researcher at MIT’s Computer Science and Intelligence laboratory. And is keen on making robots better understand the stimuli of their environment. The method Neuman uses to deal with this mind-body disagreement and delay is called robomorphic computing.
Understanding robomorphic computing
Now robomorphic computing is sure to help produce customized brains for robots. It uses the combo of the physical design of the robot and the application that needs to be worked upon. And to produce a personalized chip that reduces the robot’s response time to the stimuli. Once built, such customized brains for robots shall prove to be a state-of-the-art technology for core frontline fields like that of medical. People like medical practitioners and doctors are the frontline attendants who deal with all kinds of patients including contagious ailments.
How does a robomorphic model work?
As scientists believe that every robot is different and is built to cater to different sets of needs. Thus, they intend to achieve customized brains for robots through robomorphic computing. And hence a robot’s computational speed, accuracy, response time need to be a lot personalized to achieve the intended quick results. The robomorphic model accepts various requirements as inputs from the user. Now, these can be limbs’ design and their movement specifications, which are then converted into sparse mathematical matrices. That is these matrices can have zero values, which relate to the restricted limb movements of the robots (depending on its design layout). So, based on all of these parameters and inputs, a customized physical design and hence a customized brain for robot are prepared.
Steps involved in robot’s operation
Neuman clearly explains the steps and sequences involved in a particular robot’s operation of its understanding and responding to a particular situation. First robots gather their surrounding data with their sensors and cameras. This is called perception. Next is the mapping of this data and localization of the robot itself into its surroundings. In this step, robots fabricate a map of its surroundings and make themselves a part of it. And then most importantly, rolling out a plan and executing it. This is where the amount of time required for computing and executing these steps comes into the picture. Plancher throws light on the scenario about how increasing the computational speed is the demand of the dynamic environment. But to achieve faster processing and reduce the robot’s response time significantly, improvising only software isn’t enough. And thus, requires a holistic approach that focuses on enhancing the hardware coupled with great software.
Hardware acceleration-GPU
A GPU or Graphical Processing unit is a good example of parallel and swift processing. And is greatly known for its speed acceleration ability. A GPU’s performance is immensely fast for any specific application.
Promising results– Customized brains for robots
Hardware designs made with the help of robomorphic computing surpassed the designs built on conventional CPU and GPU systems. Though Neuman and her team worked only on programming of a customizable chip known as FPGA i.e. field-programmable gate array. And even after having a little sluggish clock rate, the robomorphic chip was still 86 times quicker in response than the conventional GPU and 8 times quicker than CPU.
Also Read: New technology to investigate ASD
Robomorphic Computing-the future:
Newman, Plancher, and all other scientists are delighted and hopeful with results of the robomorphic computing. The team is optimistic that the next 20 years could see a sharp surge in customized robots regaling with a wider set of audience. And help them fight situations that involve direct human encounters. There they also wish to automate the process of robomorphic computing entirely with simple drag and drop functionality. This shall further ease out the process for users to create personalized robots and customized brains for robots.
Authors and co-authors of the research:
Along with Neuman, Thomas Bourgeat, Srini Devadas, Brian Plancher, Thierry Tambe and Vijay Janapa Reddi have contributed towards bringing out the process of robomorphic computing.
