Section: New Results

Reasoning for relaxing traffic regulations

Participants : Philippe Morignot, Fawzi Nashashibi.

This work [39] deals with relaxation of traffic rules in unusual but practical situations. For example, if a truck is unloading on a roadway, the automated vehicle should overtake it despite a continuous yellow line: traffic rules are indeed broken, which is illegal, but this might be tolerated due to the unusual aspect of the situation at hand.

An ontology has been developed in order to represent the road network (a directed graph, vertices being intersections and edges being lanes), the infrastructure (road signs, marks), the other road users and the intelligent vehicle. Reasoning on this representation is performed by inference rules (IF/THEN symbolic structures), encoding the deliberation on the encoded situation. Main rules conclude on the next discrete motion of the vehicle, e.g., “pass onto the adjacent lane” which involves crossing a continuous yellow line.

In practice, this ontology has been created using the PROTEGE ontology editor from Stanford University. IF/THEN rules are represented in SWRL (Semantic Web Rule Language), using the reasoner PELLET from the company Clark & Parsia (a plug-in of the tool PROTEGE).

Work over the next year involves porting this reasoning module on the vehicles: porting the generated JAVA source code as one component inside the RTMAPS architecture of CyberCars, and linking the ABoxes (assertional boxes) of the ontology to symbols extracted from signals by perception.