Motional has released the Motional dataset called nuReasoning, an open dataset designed to help autonomous vehicles handle complex driving situations with more human-like reasoning and decision-making. The dataset contains more than 20,000 long-tail driving scenarios and 247,000 human-verified reasoning annotations, giving researchers detailed information about why particular driving decisions are made.
The new dataset is aimed at improving autonomous driving systems in situations that are difficult to predict, including unusual pedestrian activity, road construction, animal crossings, poor visibility, and other rare events. Motional developed nuReasoning with the UCLA Mobility Lab and Professor Jiaqi Ma.
What Is the Motional Dataset?
The Motional dataset, known as nuReasoning, is designed to help autonomous vehicles understand not only what is happening around them but also why a specific driving decision may be appropriate.
The dataset includes more than 105 hours of selected real-world driving scenarios collected from Motional operations in Las Vegas, Pittsburgh, Los Angeles, Boston, and Singapore. Each scenario contains video and detailed reasoning annotations that explain driving decisions and possible alternatives.
How Motional Dataset Supports Autonomous Driving
Autonomous vehicles must respond safely to situations that may not occur frequently but can create serious risks. Motional says nuReasoning provides training material covering spatial reasoning, decision reasoning, and counterfactual reasoning.
For example, one scenario shows a vehicle stopping near a nighttime construction area. The detailed annotation explains that the vehicle stopped because a small animal was crossing the road, while alternative routes were considered unsafe because of construction barriers and uncertainty about the animal’s movement.
This type of information can help researchers study how vehicles interpret difficult situations and improve planning and decision-making.
Vehicle Motion Data From Real-World Driving
The vehicle motion data behind nuReasoning comes from millions of miles of Motional driving data. The dataset combines multiple sensor types to create a detailed representation of road environments.
With 247,000 reasoning annotations and more than 105 hours of selected scenarios, the vehicle motion dataset gives researchers a large collection of challenging situations to study. Motional has also integrated its Omnitag search system, allowing researchers to find scenarios by location, difficulty, scenario type, and natural-language descriptions.
Motional Autonomous Driving Research Expands
The release builds on Motional’s previous open datasets, including nuScenes, nuImages, Panoptic nuScenes, and nuPlan. The company says these resources are intended to support research into vehicle perception, trajectory planning, safety, and explainable decision-making.
Motional is also launching the nuReasoning Challenge with the UCLA Mobility Lab at the European Conference on Computer Vision in Sweden. The challenge will test planning and reasoning across 1,000 private scenarios, with winners expected to be announced at NeurIPS in December.
What This Means for Autonomous Vehicles
The Motional dataset represents a broader effort to give researchers access to detailed driving examples that go beyond basic vehicle movement. By opening complex scenarios and reasoning annotations to the research community, Motional is encouraging further work on safer and more explainable autonomous driving.
For researchers searching for a Motional GitHub resource or related vehicle motion data, the nuReasoning release provides a significant new reference point for studying complex driving behavior and decision-making.
The Road Ahead for Motional
Motional’s latest dataset highlights how advanced autonomous driving research is moving toward systems that can better understand unusual road situations and explain their decisions. As researchers continue testing nuReasoning, its real-world scenarios could contribute to safer planning and more reliable autonomous vehicle technology.
Read our latest interview with Anthony Monteiro
