State Complexity Evaluation of Autonomous System using Deep Neural Networks
Abstract
Robots in Manufacturing Systems
Robots are widely used in manufacturing systems to deliver complex operations with ease. Robots are capable of introducing maximum efficiency, safety and competitive advantage [1] with automation of repetitive tasks, reduction of marginal error to negligible values and complimenting human focus on productive areas of manufacturing. Fully autonomous robots offer high volume tasks management with speed, accuracy and durability.
Path Failure Issues in Robotic Systems [2]
One might be amazed with the promises and delivery of robots in manufacturing operations and would like to know how robots deliver each process with high accuracy. Motion planning is the key to robot success. It's a term used in robotics which enables a robot in a sequential manner with configurations so that it can reach from source to destination. This paper explores the solutions to offer advantages of robotics in manufacturing with minimized path failures.
Solutions to Path Failure Issues
The key reasons that account for robotic path failures are Time Delay, Closely programmed points and inadequate system requirements at scale in production. Technologies like artificial intelligence are a great hope today for manufacturing industries to come up with self managed robots. But, currently available complexity metrics evaluating robot's programming code are inefficient in delivering a real-time self managed robot. As per detailed literature exploration [3] common actions to overcome path failures are reduction of number of instructions between consecutive move instructions and Utilization of broader space points for motion planning, reduction of speed to manage early/delay time issues.
Proposed Methods [4] for Solving Path Failure Issues [5] using Artificial Neural Networks
This paper proposes a new state complexity metrics for evaluation of path failures [6] in real time by deep learning the features of prefetch time and interpolation buffer startup adjustment using an artificial neural network, hereby protecting and preventing the failures by reconstructing the previous states and using the previous states for prediction of next states with better weights.