Skip to 0 minutes and 5 secondsThere is a nice analogy between autonomous cars and the way we, as humans, we process information and the way we navigate. We have two eyes, and the video camera the autonomous car needs to have, also, has at least two cameras in order to be able to find distances to people and other cars. We have visual information, we have sound, we have LIDAR, which can build a map of the environment, then we combine everything. So the car is autonomous when it can use that information and make decisions independently. We are developing intelligent algorithms, sensor data fusion, which extracts the meaningful information from big data. These are machine learning algorithms that help operators and also give autonomy to the car.
Skip to 1 minute and 8 secondsThere are lots of challenges. One of them is the big volume of data. We need to extract only the meaningful and build up learning algorithms. Because you can think about occlusions, changeable environments including snow, fog - that's part of one of the big challenges with autonomous cars. They need to be able to avoid collisions and detect obstacles in front of them. This video shows a machine learning algorithm which is able to detect different kinds of events and warn the driver, or the autonomous car, that there is something happening in front of it. The car has to be able to adjust to that environment.
Skip to 2 minutes and 1 secondThis is part of a video surveillance system where people are moving, and the purpose of this system is to detect pedestrians and track their motion, and possibly analyse behaviour. One application area for autonomous vehicles is airports and train stations. You can think about Heathrow, Terminal Five, where the pods are transporting passengers from the terminal to the parking place. The pods, they need to know how many passengers are at Terminal Five and how many they need to transport to the parking. And with my research, we focus on sensor data fusion algorithms that can detect the number of people that are at the parking, and then the pods will transport them to the terminal.
Skip to 2 minutes and 57 secondsBefore we see autonomous cars being ubiquitous there's a long way to go. Whereas it's much easier to have connected vehicles, because the vehicle can have all kinds of needed sensors and be connected with other vehicles, and with a transportation centre. In order to facilitate commuters to go from place A to B, we are developing algorithms that can predict better paths, and the shortest, or the most beautiful paths. Or going through areas where there are historical monuments, the most economical paths, and those with least pollution. So these are different objectives which can be achieved. We have data from the city centre of Birmingham. These are traffic loop data. We have video camera feeds.
Skip to 3 minutes and 56 secondsWe have a GSM data, possibly data from bus stations, taxi drivers, and others, and from metropolitan areas. One of the objectives is to predict the traffic. Also we have social networks. So we have the involvement of social network information from Twitter, from Facebook. If people want to find out in real time what's the weather, what's the traffic status, they can use Twitter and online data, then send their information, and improve their mobility. We are working together with the city centres. It depends on how quickly the cities, the councils, would accept this technology in our everyday life. It should come in the next two, three years.
Processing sensor data
Autonomous systems need to process large volumes of data which can arrive from multiple, diverse sensors such as RADAR, LIDAR and cameras. In this video, Dr Lyudmila Mihaylova explains how her research is tackling this challenge, by developing intelligent algorithms that can extract meaningful information from these large data sets.
Lyudmila also discusses her research with the SETA consortium, which is helping to change the way that mobility is organised, monitored and planned in large metropolitan areas. By collecting and processing dynamic data from people, cars, city sensors and distributed databases, this project hopes to inform decision makers on how to improve town planning and infrastructure, as well as allowing individuals to plan their journeys in a more efficient and sustainable way.
Would you like to see this sort of technology applied to where you live? How do you think it could help you?
© The University of Sheffield