Many robotic applications require autonomous decision making. The obstacles may be uncertain in nature. Because of the mobility, most robots might be battery operated. This chapter briefs the energy and performance analysis of robotic applications using artificial neural networks. This chapter is designed to understand the operation of robots from understanding sensor data (training), processing (testing) the data in an efficient manner, and respond (prediction) to the dynamic situation using self-learning and adaptability.
TopIntroduction To Artificial Neural Network
Human brain study has a long history. In advancement of technology, it’s rational in attempting and managing the style of thinking. ANN was established in late 1943 with the publication of a study on the potential uses of neurons by Warren McCulloch (neurophysiologist) and Walter Pitts (mathematician) (ANNs). They created a subtle model using electric circuits. Thanks to the amazing capability ofinferring the meaning of complex and irrelevant data, neural networks can be used to uncover patterns and discern trends from data that is too complex for either people or other computer systems to pick up on. A trained neural network could be considered an “expert” in the area of the data to be analysed.
A different approach of solving a problem by using an algorithmic approach, Traditional computers handle issues by following a set of instructions, which includes the computer. The computer is unable to resolve the issue without knowing the precise procedures that must be followed.Because of this, only issues that humans can currently understand and be familiar with may be solved by conventional computers. However, computers would be so much more useful if they could carry out activities that we are unclear of how to do. Neural networks process information in a manner akin to that of the human brain. The network is made up of many processing units that are closely connected to one another and work together to solve a certain problem simultaneously. Neural networks can learn from examples. They can't be forced to complete a pre-planned assignment. To avoid wasting time or, worse yet, having the network behave poorly, the examples must be carefully picked.
The network determines on fixing the problem by itself, its behaviour is unpredictable. Traditional computers, use a cognitive approach in problem solving; the answer is known and provided in a few simple, understandable steps. Once these instructions have been converted into a high-level language programme, the computer can comprehend them. These devices will only have hardware or software problems, if any.
These devices are entirely foreseeable. The usage of classic algorithmic computers and neural networks complements rather than competes with each other. There are tasks that are suited for neural networks, whereas arithmetic operations, are suited for algorithmic approaches. In addition, many activities require a combination of the two techniques in order for systems to operate at their best (often, a traditional computers require to oversee the neural network) (Maind et.al, 2014).
ANN: Characteristics
The characteristics of ANN are as follows.
Parallel Processing Capability
ANN is only beginning to learn about the concept of parallel preparing in the realm of computers. Human neurons perform parallel processing, which is incredibly unpredictable, but by using basic parallel preparation techniques, like matrix and some lattice estimations, we can mimic this in ANN.
Distributed Memory
We must therefore store data atweight lattice, sort of long-lasting memory, data is stored as examples throughout the system arrangement because ANN is a very large system and particular location memory or unified memory cannot satisfy ANN system's needs.
Capability for Fault Tolerance
Because ANN is a very complicated system, fault tolerance is a must. Because even if one component fails, the system as a whole won't be affected as much; nevertheless, if every component fails at once, the system will entirely malfunction.
Collective Solution
Because ANN is a networked system, the system's result is a summation of total inputs it has processed, or a combined output of those inputs. The incomplete response is useless to any ANN System user.
Learning Capability
Most learning rules used in ANNs are used to simulate processes, adjust the network to its changing environment, and gather useful data. Unsupervised, supervised, and reinforcement are these learning techniques.