Читать книгу AI and IoT-Based Intelligent Automation in Robotics - Группа авторов - Страница 31

Artificial intelligence works based on some algorithms and those algorithms are studied under machine learning (ML). These algorithms are used based on the requirement of the type of task to be carried out and the final goal to be achieved. ML algorithms are categorized into the four basic types depicted in Figure 2.8 below.

The algorithms that fall under each category have been derived on the basis of mathematical and statistical inferences and work as per some mathematical model—everything can be depicted in terms of mathematical models and those models form the basis of machine learning algorithms that are to be used in robotics. In some industrial robots, the motion planning of robotic arms is done via algorithms such as the Bayesian filter. The Bayes’ rule has some fascinating roles in robotics that is hidden under a single equation:

Figure 2.8 Types of machine learning algorithms.

The above expression can be understood as:

 Expression before the integral can be understood as: make a guess and improve it by reading the sensor data.

 Expression within integral or after integral can be understood as: draw what we already know and try to guess to make it better.

The above algorithm can be used to derive some other algorithms like:

1 Algorithms for linear and non-linear systems:Linear:− Linear Kalman filterNon-linear:− Extended Kalman filter− Unscented Kalman filter

2 Improved version of Kalman filter: Information filter

3 Particle filter: Used in the Monte Carlo method

4 Histogram filter: For making multidimensional items and histograms

For more algorithms used in robotics please refer to [21, 22].