This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A Jiushi autonomous delivery vehicle was filmed dragging a fallen electric scooter in Xianyang, China, after failing to detect the obstacle. The vehicle continued to drive despite the wheel being jammed by the scooter. The manufacturer acknowledged the incident, citing a need to improve the algorithm for detecting irregularly shaped obstacles, and confirmed no injuries occurred.
A Jiushi autonomous delivery vehicle reportedly operating on a delivery route in Xianyang, Shaanxi, was filmed dragging a fallen electric scooter that had allegedly been left in the road after an earlier traffic incident. Company statements said the vehicle attempted to avoid the scooter but failed to fully recognize its shape and position, causing part of it to become caught under a wheel. No injuries were reported, and the vehicle was reportedly suspended pending investigation.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that fail to perform reliably or effectively under varying conditions, exposing them to errors and failures that can have significant consequences, especially in critical applications or areas that require moral reasoning.
The autonomous vehicle failed to detect a fallen scooter due to limitations in its perception system's ability to recognize irregularly shaped obstacles, and its control system lacked the robustness to detect a locked wheel.
AI system
Due to a decision or action made by an AI system
Unintentional
Due to an unexpected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The incident was caused by the autonomous delivery vehicle's perception and control system failing to detect a fallen scooter and continuing to drive after the collision.
An autonomous delivery vehicle in China failed to detect and drag a fallen electric scooter due to a perception algorithm limitation. The incident was a minor, unintentional commercial traffic accident with negligible national security implications across all assessed categories.
The delivery vehicle operates independently on public roads, navigating traffic lights, pedestrians, and obstacles using cameras and sensors without real-time human intervention.
High Risk: The autonomous delivery vehicle operates in public transport and traffic environments, which represents an application in critical infrastructure with significant implications for public safety.
The AI system's primary function is to autonomously navigate urban streets to deliver packages without a human driver.
The reports provide consistent details about the primary incident, including timestamps from the vehicle's log, the manufacturer's official response, and the specific technical failure.
None. The event is clearly a perception and control failure of an autonomous delivery vehicle.
Technology
Perception Shape Recognition Limit
Perception system struggled to identify side-lying electric scooter shape.
Locked Wheel Detection Failure
The system did not detect or halt when the left front wheel locked up.
Improper Post-Collision Acceleration
Vehicle resumed normal speed after detecting collision and slowing down.
Data Inputs
Missing Side-lying Vehicle Data
Lack of training data for side-lying scooters led to classification error.
Inadequate Obstacle Shape Labels
Algorithm lacks precise labeling for irregularly shaped road debris.
Human Factors
Delayed Remote Operator Intervention
Operators took minutes to manually pause the vehicle after the collision.
Operator Remote Control Error
Remote operator rear-ended a parked SUV in manual mode the next day.
Process and Methods
Insufficient Scenario Testing
Lack of multi-dimensional scenario testing for weirdly-shaped obstacles.
Inadequate Post-Collision Protocol
Safety protocols allowed the vehicle to resume speed after a collision.
Management
Prioritizing Deployment Over Safety
Accelerated commercial deployment in 20 cities despite software limits.
Inadequate Risk Assessment
Company failed to assess risks of locked wheels and abnormal obstacles.
Regulatory Environment
Rapid Commercial Adoption Incentives
Government incentives accelerated rollout before mature safety regulations.
Lack of National Safety Standards
Absence of strict national standards for autonomous vehicle fault handling.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
