Can intelligent driving systems pass in extremely narrow lanes when meeting an oncoming vehicle at close range?
Video
Summarizing, urban villages have a few typical characteristics:
High density and low ratio: high population density, high building density, and low floor area ratio.
Self-contained traffic systems: lack of internal road networks, with "no rules" being the only rule.
Complete miniature life and commerce circle: each urban village is a compact, fully functional mini life and business hub.
Video↓
Urban Villages in the Eyes of Autonomous Driving: Traffic Rules Degrade, and Every Meter is a Long-Tail Scenario
In the eyes of autonomous vehicles, urban villages are full of dynamic obstacles, with no recognizable traffic signs and extreme, unpredictable challenge scenarios. Typical long-tail scenarios commonly encountered by autonomous vehicles in urban villages include:
Uncertain obstacle paths: Without lane markings or traffic lights, the roads are filled with cars, bicycles, electric scooters, and pedestrians moving unpredictably in both speed and direction.
Close-range interactions, brushing past vehicles and pedestrians: With extremely narrow roads and insufficient passing space, coupled with a high density of pedestrians, autonomous vehicles must interact within centimeters of surrounding cars and pedestrians, often needing to squeeze past oncoming vehicles.
Small obstacle challenges in blind spots: Small obstacles such as dogs or scattered goods frequently appear on the road, significantly increasing the risk of blind spots for autonomous vehicles.
Huawei has consistently used a universal autonomous driving algorithm to tackle all application scenarios and continuously evolve through learning. Now, Huawei's system integrates algorithms that incorporate environmental observation, bidirectional negotiation, consensus-building, and mutual understanding on top of perception, prediction, planning, decision-making, and control, achieving safe and stable autonomous driving. In solving the complex traffic issues of urban villages, Huawei’s approach has three major pillars: embracing uncertainty, building consensus, and overcoming blind spots.
Embracing Uncertainty: More Uncertainty, More Stability
Urban villages’ dynamic and unpredictable traffic challenges Huawei's predictive algorithms. Huawei responds to this by embracing uncertainty in prediction.
Huawei strengthens its predictive models with machine learning. These models are trained on large datasets to extract obstacle movement patterns, continuously enhancing the accuracy of predictions.
One notable feature of Huawei's model is its adaptive interaction logic. When autonomous vehicles interact with nearby cars and pedestrians, the prediction is updated in real-time, based on each interaction step, rather than relying solely on fixed-time prediction data. By tightly connecting prediction with planning, decision-making, and control stages, the system achieves greater stability, resulting in smoother vehicle operation.
Building Consensus: Enhancing the Vehicle’s Proactivity in Human-Machine Negotiation
With degraded traffic rules in urban villages, traffic relies on "consensus" among vehicles and pedestrians. How can machines "reach a consensus" with their surroundings? Autonomous vehicles communicate through actions.
Under the principle of safety first, Huawei’s autonomous vehicles prioritize yielding to pedestrians and other vehicles. In dense urban village streets, yielding might cause the vehicle to halt indefinitely. Huawei’s vehicles take proactive actions by adjusting direction and speed to signal surrounding pedestrians and vehicles, seeking mutual understanding.
Building consensus is central to addressing urban village traffic and is a key technical focus for Huawei.
Overcoming Blind Spots: Long-Range Visibility, Close-Range Precision
Huawei’s self-developed autonomous driving sensor suite detects all road and traffic information within a 250-meter range, providing 360-degree coverage of obstacles. This precision allows Huawei’s vehicles to reliably “see” surrounding objects, distinguishing, tracking, and understanding scenes.
However, in the crowded environment of urban villages, close-range detection poses significant challenges. Huawei’s autonomous vehicles often move within centimeters of obstacles, requiring near-zero blind spots.
For urban villages, Huawei upgraded its sensor suite with high-density side-laser LiDAR, enhancing detailed detection of nearby obstacles and greatly improving detection accuracy.
03 Moving Toward Fully Autonomous Driving
Developing Robotaxi services for urban traffic requires the ability to handle a wide variety of complex traffic issues on open city roads. One of the biggest challenges for autonomous driving is the long-tail scenarios involving non-compliant vehicles or pedestrians behaving unpredictably. Huawei accelerates the advancement of autonomous driving through intensive long-tail scenario training and algorithm iteration.
Solving autonomous driving issues in urban villages not only enables safer, smoother, and more stable transportation services on open city roads but also accelerates the removal of safety drivers, moving toward fully autonomous operations.
Large differences inside and outside: often located around bustling CBDs, urban villages have vastly different traffic conditions inside and outside.
In China’s commercial areas, building density rarely exceeds 40%-50%, while some urban villages reach up to 70%. Urban villages typically have only one or two external roads of around 7 meters wide, with internal roads mainly formed by the spacing between buildings, usually 2-4 meters wide, accommodating both pedestrians and vehicles without lane markings or traffic lights. Private vehicles and street vendors occupy much of the road space, leaving narrow paths for passing cars. However, upon exiting the village, broad streets near subway stations and commercial districts contrast sharply with the village’s internal conditions, creating a marked difference between the inside and outside traffic scenarios.
Summarizing, urban villages have a few typical characteristics:
High density and low ratio: high population density, high building density, and low floor area ratio.
Self-contained traffic systems: lack of internal road networks, with "no rules" being the only rule.
Complete miniature life and commerce circle: each urban village is a compact, fully functional mini life and business hub.
02 Urban Villages in the Eyes of Autonomous Driving: Traffic Rules Degrade, and Every Meter is a Long-Tail Scenario
In the eyes of autonomous vehicles, urban villages are full of dynamic obstacles, with no recognizable traffic signs and extreme, unpredictable challenge scenarios. Typical long-tail scenarios commonly encountered by autonomous vehicles in urban villages include:
Uncertain obstacle paths: Without lane markings or traffic lights, the roads are filled with cars, bicycles, electric scooters, and pedestrians moving unpredictably in both speed and direction.
Close-range interactions, brushing past vehicles and pedestrians: With extremely narrow roads and insufficient passing space, coupled with a high density of pedestrians, autonomous vehicles must interact within centimeters of surrounding cars and pedestrians, often needing to squeeze past oncoming vehicles.
Small obstacle challenges in blind spots: Small obstacles such as dogs or scattered goods frequently appear on the road, significantly increasing the risk of blind spots for autonomous vehicles.
The new Chinese intelligent driving system has consistently used a universal autonomous driving algorithm to tackle all application scenarios and continuously evolve through learning. Now, this system integrates algorithms that incorporate environmental observation, bidirectional negotiation, consensus-building, and mutual understanding on top of perception, prediction, planning, decision-making, and control, achieving safe and stable autonomous driving. In solving the complex traffic issues of urban villages, the new Chinese intelligent driving system approach has three major pillars: embracing uncertainty, building consensus, and overcoming blind spots.
Embracing Uncertainty: More Uncertainty, More Stability
Urban villages’ dynamic and unpredictable traffic challenges the system’s predictive algorithms. The new Chinese intelligent driving system responds to this by embracing uncertainty in prediction.
The system strengthens its predictive models with machine learning. These models are trained on large datasets to extract obstacle movement patterns, continuously enhancing the accuracy of predictions.
One notable feature of this system's model is its adaptive interaction logic. When autonomous vehicles interact with nearby cars and pedestrians, the prediction is updated in real-time, based on each interaction step, rather than relying solely on fixed-time prediction data. By tightly connecting prediction with planning, decision-making, and control stages, the system achieves greater stability, resulting in smoother vehicle operation.
Building Consensus: Enhancing the Vehicle’s Proactivity in Human-Machine Negotiation
With degraded traffic rules in urban villages, traffic relies on "consensus" among vehicles and pedestrians. How can machines "reach a consensus" with their surroundings? Autonomous vehicles communicate through actions.
Under the principle of safety first, the new Chinese intelligent driving vehicles prioritize yielding to pedestrians and other vehicles. In dense urban village streets, yielding might cause the vehicle to halt indefinitely. The system’s vehicles take proactive actions by adjusting direction and speed to signal surrounding pedestrians and vehicles, seeking mutual understanding.
Building consensus is central to addressing urban village traffic and is a key technical focus for the new Chinese intelligent driving system.
Overcoming Blind Spots: Long-Range Visibility, Close-Range Precision
The new Chinese intelligent driving system’s self-developed sensor suite detects all road and traffic information within a 250-meter range, providing 360-degree coverage of obstacles. This precision allows vehicles to reliably “see” surrounding objects, distinguishing, tracking, and understanding scenes.
However, in the crowded environment of urban villages, close-range detection poses significant challenges. Vehicles often move within centimeters of obstacles, requiring near-zero blind spots.
For urban villages, the system upgraded its sensor suite with high-density side-laser LiDAR, enhancing detailed detection of nearby obstacles and greatly improving detection accuracy.
03 Moving Toward Fully Autonomous Driving
Developing Robotaxi services for urban traffic requires the ability to handle a wide variety of complex traffic issues on open city roads. One of the biggest challenges for autonomous driving is the long-tail scenarios involving non-compliant vehicles or pedestrians behaving unpredictably. The new Chinese intelligent driving system accelerates the advancement of autonomous driving through intensive long-tail scenario training and algorithm iteration.
Solving autonomous driving issues in urban villages not only enables safer, smoother, and more stable transportation services on open city roads but also accelerates the removal of safety drivers, moving toward fully autonomous operations.
Large differences inside and outside: often located around bustling CBDs, urban villages have vastly different traffic conditions inside and outside.
In China’s commercial areas, building density rarely exceeds 40%-50%, while some urban villages reach up to 70%. Urban villages typically have only one or two external roads of around 7 meters wide, with internal roads mainly formed by the spacing between buildings, usually 2-4 meters wide, accommodating both pedestrians and vehicles without lane markings or traffic lights. Private vehicles and street vendors occupy much of the road space, leaving narrow paths for passing cars. However, upon exiting the village, broad streets near subway stations and commercial districts contrast sharply with the village’s internal conditions, creating a marked difference between the inside and outside traffic scenarios.