Knowledge graphs (KGs) are increasingly used in autonomous driving (AD) to make traffic scene entities and relations explicit for prediction and reasoning. Existing traffic-scene KGs, however, primarily emphasize trajectory forecasting or general scene understanding. This paper presents the nuScenes Intention Knowledge Graph (NIKG), a focused Web Ontology Language (OWL)-based semantic representation and construction pipeline for ego-centric vehicle intention context. NIKG instantiates nuScenes scenes as a self-describing Turtle graph and promotes three intention-relevant relations to first-class properties of each scene participant: ego-relative motion, spatial configuration, and time to collision (TTC) risk. The pipeline uses a small input contract available in many ego-centric driving datasets: ego pose, tracked object annotations, and temporal keyframe order. NIKG is released as an ontology-backed, self-describing graph artifact together with a reproducible construction pipeline and validation protocol for ego-centric intention context. Applying this pipeline to nuScenes mini yields a graph of 211,335 triples across 404 keyframes; under the validation protocol, ego-relative motion labels achieve 92.5% forward consistency. Coverage, semantic consistency, and temporal coherence analyses over this artifact are further reported. NIKG occupies the space between scene-oriented traffic KGs and intention-prediction models: a small, explicit semantic layer intended to support downstream work such as knowledge graph embedding (KGE), SPARQL-based feature extraction, or hybrid neuro-symbolic reasoning.
Figures & tables
Figure 1: NIKG in the autonomous driving pipeline. The semantic context layer (dashed, this work) sits between perception and the downstream intention predictor, making ego-relative motion, spatial configuration, and time to collision risk available as explicit, queryable graph facts.
Figure 2: NIKG ontology structure. Nodes represent OWL classes; edges are object properties. The three intention-relevant relations ( motionTowardEgo , spatialConfiguration , ttcRisk ) connect SceneParticipant to their respective enumeration classes. hasCoarseCategory connects the underlying Participant to AgentCoarseType .
Table 1: Intention-relevant object properties introduced by NIKG.
Figure 3: NIKG generation pipeline. Each stage is independently testable; the final artifact merges schema and instance triples into a single self-describing Turtle file.
Signal
Source
Used for
Ego pose (x,y,ψ)
ego record
ego frame + time to collision
Annotation position
annotation
participant geometry
Instance identifier
annotation
track continuity
Keyframe order
sample chain
temporal relations
Optional image/intrinsics
camera metadata
figures / future 2D
Table 2: Minimal input contract for graph instantiation.
Figure 4: Scene-1100, frames 19–22 (0.5 s apart). Bounding boxes are colour-coded by ttcRisk : red = High , orange = Medium , green = Low , grey = none. Each box label: coarse category / motionTowardEgo — spatialConfiguration — ttcRisk .
Figure 5: Representative NIKG subgraph for scene-1100 frame 22. Two SceneParticipant nodes are expanded to show their populated intention-relation triples; the remaining 21 participants in the frame are collapsed for clarity. The temporal chain linking this frame to frames 21 and 23 is visible at the top of the diagram.
Statistic
Value
Total triples
211,335
Scenes (keyframes)
404
Unique participants
911
SceneParticipant instances
18,538
hasCoarseCategory triples
18,538
hasNuScenesCategory triples
18,538
Table 3: Graph scale on nuScenes mini (10 scenes).
Constraint
Total
Pass
Rate
ttcRisk=High ⇒ Approaching
1,990
1,988
99.9%
ttcRisk=High ⇒ spatialConfiguration in forward zone
1,892
1,688
89.2%
Stationary ⇒ ttcRisk = High
3,784
3,782
99.9%
Table 4: Semantic consistency checks. Pass rate = fraction of premise instances that also satisfy the conclusion.
Knowledge graph construction turns text, tables, images, and databases into structured entities and relationships that machines can query and reason over.
Institute for Intelligent Systems, Esslingen University of Applied Sciences, Germany. · Department of Computer Science, University of Freiburg, Germany.