Towards an Intelligent Driver for an Autonomous Car
There are a number of active investigations into various aspects of autonomous driving, which are briefly summarized:
Pedestrian behavior understanding
The goal of this project is to observe and understand pedestrian actions at the time of crossing, and identify the factors that influence the way pedestrians make crossing decision. We intend to incorporate these factors into predictive models in order to improve prediction of pedestrian behavior.
For more detailed information regarding this project, please click here.
Pedestrian intention estimation
The objective of this project is to develop methods to predict underlying intention of pedestrians on the road. Understanding the intention helps distinguish between pedestrians that will potentially cross the street and the ones that will not do so, e.g. those waiting for a bus. To achieve this objective we want to establish a baseline by asking human participants to observe pedestrians under various conditions and tell us what the the intention of the pedestrians were. We want to use this information to train an intention estimation model and examine how it can improve the prediction of pedestrians’ trajectories and actions.
For more detailed information regarding this project, please click here.
Pedestrian crossing action and trajectory prediction for autonomous vehicles
Different approaches using recurrent neural networks within encoder-decoder ensembles are being designed, implemented and evaluated to predict future crossing/not crossing action of pedestrians, as well as their future trajectories both in 2D and in 3D, using a monocular camera onboard a vehicle and different features such as appearance, pedestrian location, ego-vehicle dynamics, etc. The main goal is to devise a system capable of inferring future behaviours and locations of pedestrians to improve the safety of current advanced driver assistance systems and autonomous vehicles.
This work is in collaboration with Dr. David Fernández Llorca, University of Alcalá, Alcalá de Henares (Madrid), Spain.
Lane change prediction for autonomous vehicles
Lane change prediction of surrounding vehicles using appearance, local context and optical flow features. Images from a frontal-view camera onboard of a vehicle are used as the main source of information. Several two-streams CNN-based architectures are being implemented and evaluated. The final goal is to devise a system able to infer future lane changes of surrounding vehicles for autonomous vehicles.
This work is in collaboration with Dr. David Fernández Llorca, University of Alcalá, Alcalá de Henares (Madrid), Spain.
Task-based attention for driving
This project will explore explicit modelling of driver’s attention, such as task-based saliency, and various implicit attention mechanisms common in the deep learning literature. The goal is to develop an approach which can localize and prioritize objects (or object parts) relevant for driving and potentially lead to performance improvements on visual tasks involved in driving, such as road user action prediction and object detection. Even though we do not aim for a biologically-realistic solution we plan to collect human behavioural data (in-lab and in-vehicle) that can be used for training and evaluation of the algorithms.