Intelligent mobile robots have been quite successful while collaborating and operating with humans in daily living environments and this success depends on their ability to learn and generalise human movements, and gain a shared understanding of a scene. 

Unsupervised Learning

Unsupervised learning means training a machine by using data that is not labelled or classified and enabling the algorithm to execute data without guidance. Thus, the job of the machine is to group uncategorized or unlabelled data according to patterns, similarities, and differences without any previous data training.

Unsupervised learning frameworks enable mobile robots to be more helpful, particularly while sharing an environment with humans. By eliminating humans from the learning process, these robots can directly learn from huge quantities of data available (observations); this in turn enables robots to adapt to their environment/surrounding which saves time and effort. Therefore, unsupervised learning is learning without a teacher.

How does Unsupervised Machine Learning Work?

In unsupervised learning, an Artificial Intelligence (AI) system is presented with uncategorized data and the system’s algorithms execute the data without previous training. The output is based on coded algorithms. Exposing a machine to unsupervised learning is an established method of testing the abilities of that system. Unsupervised learning algorithms can execute more complex processing tasks than a supervised learning system; however, it can also be more unpredictable. For example, a system that is trained using the unsupervised model may learn how to differentiate between dogs and cats but it may also add unsought for behaviour such as how to deal with different breeds of dogs and cats; this may end up messing things instead of keeping them in order. Unsupervised learning is particularly helpful during tasks associated with feature extraction and data mining. The objective of unsupervised learning is to discover concealed patterns, trends, or to extract desired features from data.

Unsupervised learning deals only with the input data set without any previous knowledge; therefore, there are two types of unsupervised learning models:

  • Parametric unsupervised learning: Assumes a parametric distribution of data, this means that this type of unsupervised learning assumes that the data is derived from a population that follows a specific probability distribution depending on some parameters.
  • Non-parametric unsupervised learning: Refers to the clustering of the input data set. Each cluster, depicts something about the classes and categories of the data items present in the set. This is the most common method for analyzing data and data modelling with small sample sizes.
Note: Clustering deals with the pattern or structure in a collection of unlabelled and uncategorized data. A simple definition of a cluster could be a group of objects where every member of the group is similar to the other in some way.

Applications of Unsupervised Machine Learning

Unsupervised learning assists in:

  • Anomaly detection which means finding significant data points in data collection from a dataset. 
  • Association mining, which means classifying a set of items that are present together in a dataset. This is a useful technique for basket analysis.
  • Dimensionality reduction which means reducing the number of features in a dataset and thereby enabling better data pre-processing.


ML and AI are highly complex fields and any sophisticated AI system uses a combination of various learning mechanisms and algorithms.