Labels:
deep-learning machine-learning knowledge-graphs reasoning human-ai-interaction hybrid-AI automated-driving multi-objective-decision-making
SNOW

Description

The goal of the SALVA project is to use situation awareness for planning of an autonomous shuttle, so that the shuttle itself can decide to deviate from its original path (so called virtual rails) in order to be able to handle situations such as objects blocking the road, oncoming traffic, et cetera.

Problem Context

In 2021, TNO Integrated Vehicle Safety (IVS) worked together with parties like Gemeente Helmond and Brainport Development to implement situation awareness in automated shuttle busses. This project, called OCCTAS, showed how situation awareness in the form of context characterisation and risk assessment could improve an automated vehicles' capabilities of driving autonoumsly in a safe but also comfortable manner. The project concluded that more research needed to be done before automated vehicles could safely be implemented in society. This conclusion led to the SALVA (Situation Awareness to Leave Virtual Rails) project.

Solution

The SALVA project aims to combine three facets of automated shuttles to allow the shuttle bus to autonomously deviate from its original path. This will allow it to pass objects on the road if doing so is both necessary and safe. To allow for the autonomous motion planning, firstly, the world model of the automated shuttle needs to be able to label objects that come in on its camera stream. The object detection system should return both the type of object, such as vehicles, pedestrians, garbage bins, et cetera, and include uncertainties.

Secondly, the behaviour planner of the automated shuttle should incorporate situation awareness. The shuttle bus needs information about the context in which it operates, the competence of its perception systems in the current situation, and the risk attributed to future actions. With aid from the TypeDB reasoning system, the settings for the Model Predictive Control (MPC) planner should be adjusted so the system works optimally based on these vlaues.

Finally, the MPC plans the path around the obstacle on the road, if it is safe and necessary to leave the virtual rails (i.e. the path the shuttle bus was originally going to take). The optimal planning for the alternative path should be based on potential fields that are higher in regions where the vehicle should preferably not go.

Results

Contact

  • Jan-Pieter Paardekooper, Scientist, TNO, e-mail: jan-pieter.paardekooper@tno.nl