Transfer of materials and packages between offices in multi-storeyed buildings, traditionally done by human work force is common. Can this work be done by intelligent robots that could be more systematic and less expensive? Yes, say a two-member student-team of VIT, which has designed a ‘Line-following robot’ that can climb stairs and perform the task with precision.
The student-inventors Ayush Kumar and Pallavi Bhamare, both final year student of School of Electronics Engineering, said the caterpillar robot was designed to climb stairs with the help of Artificial Neural Network (ANN)-based technology. They say that ANNs are simplified models of human nervous system that can learn an arbitrary mapping of inputs and outputs.
“We have designed the robot to learn the mapping between inputs and control decisions such as turning left, right and straight ahead from a human trainer and climb stairs smoothly,” says the young inventors. Unlike most robots which use conventional simulation algorithm, this robot, fitted with nine sensors, collects input from a large number of training data and makes controlled decisions accordingly, they pointed out.
Both the students have spent around three-and-a-half months to complete the design work. While Ayush had focused on the mechanical design of the robot, Pallavi worked on developing the ANN algorithm using the MATLAB software.
The guide and mentor Dr Mathew Mithra Noel said that the 55-cm high and 18-cm wide robot, built at a cost of `25,000, was the first-of- its-kind and was based on a supervisor model. The robot was trained to follow a white lined track. This machine is energy efficient as it is fitted with a rechargeable battery and takes about 15-cm high steps and can easily carry upto 1.5 kg of load. “This robot has already stirred international debate and interest,” he further said.
Ayush and Pallavi said the robot could be scaled up to deliver bigger volume of packages and materials in multi-storeyed environment in industry or business houses. They are also planning to patent their model and the next step for them would be to go in developing an unsupervised learning robot.