Ever increasing autonomy of machines and the need to interact with them creates challenges to ensure safe operation. Recent technical and commercial interest in increasing autonomy of vehicles has led to the integration of more sensors and actuators inside the vehicle, making them more like robots. For interaction with semi-autonomous cars, the use of these sensors could help to create new safety mechanisms. This work explores the concept of using motion tracking (i.e skeletal tracking) data gathered from the driver whilst driving to learn to classify the manoeuvre being performed. A kernel-based classifier is trained with empirically selected features based on data gathered from a Kinect V2 sensor in a controlled environment. This method shows that skeletal tracking data can be used in a driving scenario to classify manoeuvres and sets a background for further work.

The advent of autonomous vehicles comes with many questions from an ethical and technological point of view. The need for high performing controllers, which show transparency and predictability is crucial to generate trust in such systems. Popular data-driven, black box-like approaches such as deep learning and reinforcement learning are used more and more in robotics due to their ability to process large amounts of information, with outstanding performance, but raising concerns about their transparency and predictability. Model-based control approaches are still a reliable and predictable alternative, used extensively in industry but with restrictions of their own. Which of these approaches is preferable is difficult to assess as they are rarely directly compared with each other for the same task, especially for autonomous vehicles. Here we compare two popular approaches for control synthesis, model-based control i.e. Model Predictive Controller (MPC), and datadriven control i.e. Reinforcement Learning (RL) for a lane keeping task with speed limit for an autonomous vehicle; controllers were to take control after a human driver had departed lanes or gone above the speed limit. We report the differences between both control approaches from analysis, architecture, synthesis, tuning and deployment and compare performance, taking overall benefits and difficulties of each control approach into account.

Machines with high levels of autonomy such as robots and our growing need to interact with them creates challenges to ensure safe operation. The recent interest to create autonomous vehicles through the integration of control and decision-making systems makes such vehicles robots too. We therefore applied estimation and decision-making mechanisms currently investigated for human-robot interaction to humanvehicle interaction. In other words, we define the vehicle as an autonomous agent with which the human driver interacts, and focus on understanding the human intentions and decisionmaking processes. These are then integrated into the robot's/vehicle's own control and decision-making system not only to understand human behaviour while it occurs but to predict the next actions. To obtain knowledge about the human's intentions, this work relies heavily on the use of motion tracking data (i.e. skeletal tracking, body posture) gathered from drivers whilst driving. We use a data-driven approach to both classify current driving manoeuvres and predict future manoeuvres, by using a fixed prediction window and augmenting a standard set of manoeuvres. Results are validated against drivers of different sizes, seat preferences and levels of driving expertise to evaluate the robustness of the methods; precision and recall metrics higher than 95% for manoeuvre classification and 90% for manoeuvre prediction with time-windows of up to 1.3 seconds are obtained. The idea of prediction adds a highly novel aspect to human-robot/human-vehicle interaction, allowing for decision and control at a later point.

No abstract

No abstract