How about talking to a robot and getting a real response immediately? Getting all the work done in time without breaks, holidays, and sick leaves with mistakes a rarity?
Heya! “Wish granted”. Experts have found that although robots operate quickly yet in complex situations the robots find it hard to respond swiftly. Sabrina Neuman, a Ph.D. from the MIT Computer Science and Artificial Intelligence Laboratory (CSAIL) and a postdoctoral NSF Computing Innovation Fellow at Harvard’s School of Engineering and Applied Sciences found that the robot takes much longer time in interpreting the stimulus and generating a response due to a lengthy process of computation. Thankfully, Neuman has sorted out this problem too. According to her, robomorphic computing is a process by which a robot’s physical layout and intended applications are used to create a custom-made computer chip that reduces the robot’s reaction time.
The good news is that such a type of advancement can increase the capacity of robots to be used in several ways most importantly their utility can be felt in the healthcare sector to treat contagious patients directly. Neuman affirms this and says that “It would be fantastic if we could have robots that could help reduce the risk for patients and hospital workers.”
Neuman is expected to present her findings at April’s International Conference on Architectural Support for Programming Languages and Operating Systems. MIT co-authors include graduate student Thomas Bourgeat and Srini Devadas, the Edwin Sibley Webster Professor of Electrical Engineering, and Neuman’s Ph.D. advisor. Other co-authors include Brian Plancher, Thierry Tambe, and Vijay Janapa Reddi, all of Harvard University.
Furthermore, Neuman says a robot operates in three steps. Firstly, the perception involves using sensors and cameras to collect data. Secondly, mapping and localization allow a robot to use this data, create a map of the world and localize themselves within that map. Lastly, motion planning and control take place which helps to decide on a course of action. The problem part is that these steps are time-consuming as they involve difficult computations. Plancher asserts that for robots to function in changing environments, they need to be equipped with algorithms that can help them to respond in-no-time. However, for this to happen the CPU hardware need to be updated which implies taking one step further and going for hardware acceleration to give an instant boost to a robot’s brain.
Hardware acceleration is defined as using a specialized hardware unit to perform certain computing tasks more effectively and efficiently. This can be achieved via using a graphics processing unit (GPU) which is a chip specialized for parallel processing. The system utilizing GPU can meet the computing requirements of a robot. The user of the system gives input regarding the different parameters of the robot and how its joints can move. These physical properties are then converted into mathematical matrices consisting of many zero values. The system then creates a hardware architecture especially made to do calculations of the non-zero values in the matrices. The chip made as a result of this customized process successfully meets the computing requirements of the robot as per the initial testing. Surprisingly, a chip programmed in this way works eight times faster than the CPU and 86 times faster than the GPU.
Therefore, robomorphic computing holds great potential for widening the use and application of robots as it can relieve humans from getting exposed to risky environments, for example, attending to COVID-19 patients or working with heavy objects.