Robotic vacuum cleaners and data collection
Robotic vacuum cleaners (robovacs) are compact, computerised vacuum cleaners that guide themselves around rooms and collect dust and dirt from floors.
They are typically used in homes and offices as a way of automating vacuuming and can be operated with smartphone apps.
Robovacs are smart devices, trained using machine learning. They have several sensors that help them detect dirt and avoid potential obstacles such as walls, rug tassels or wires. Once a robovac has cleaned the floors, it will return to its charging dock.
Some robovacs clean floors at random, which means that they do not build maps of the floor layout. However, others will use mapping methods, including LiDAR mapping (an airborne mapping technique that uses a scanning laser to measure the height of the terrain) or camera-based mapping such as Visual Simultaneous Localisation and Mapping (vSLAM). Robovacs that use vSLAM use inbuilt cameras to capture images of rooms and build a map of the route around the rooms.
What is a smart device?
A smart device is a device that is connected to the internet and uses communication hardware such as sensors and processors. It can collect and transmit data to other devices, allowing them to automate tasks. Typical examples of smart devices include smartphones, smartwatches, smart thermostats and virtual assistants such as Amazon’s Alexa. Smart devices can interact with each other. For instance, a virtual assistant can adjust heating levels by interacting with a smart thermostat.
What is machine learning?
Machine learning is a kind of AI. By learning from existing data (known as‘training data’), systems can identify patterns to make predictions or calculate probabilities. These systems are able to continuously adjust as they encounter more data. Based on what they learn from training data, they can perform a variety of functions, like playing chess, recognising faces or assessing welfare benefit applications.
What are the benefits of this technology?
A key benefit is that robovacs automate vacuum cleaning, saving time and providing an alternative for people who may struggle to use a conventional vacuum cleaner, such as those with mobility challenges.
What are the risks of this technology?
Robovacs that use mapping methods have the potential to violate user privacy.
As noted, some models of robovacs are equipped with vSLAM technology. Recently, images of users captured by some robovacs were leaked online. In this case, users were testing the vacuums, and had signed an agreement that recordings would be used for training. And while manufacturers state that they are committed to safeguarding the privacy of personal data, the training incident highlights the potential risks of privacy violations.
Additionally, robovacs use machine learning, which is data intensive, and the labour of data annotation is often outsourced to low-paid workers. Data annotation is the process of labelling the data that a machine learning model is trained on. It is a way of tagging the information contained in a dataset (that could include images, videos or text) so that it can be fed into the machine learning model, which will subsequently be able to identify patterns in the data. The outsourcing of data annotation raises concerns around worker exploitation.