This blog post is predicated on the keynote speech within the End-to-end Autonomous Driving Workshop at CVPR 2023 held in Vancouver, titled “The Practice of Mass Production Autonomous Driving in China”.
Autonomous driving is a frightening challenge, especially in China, where human driving is already one of the difficult on this planet. There are three predominant aspects that comes into play: dynamic traffic participants, static road structure, and traffic signals. Particularly, traffic light control signals pose a singular challenge as they’re . In the next sessions, we are going to review the dynamic objects and static environments briefly, and do a deep dive on the interesting and special topic of traffic light.
Dynamic traffic participants, resembling vulnerable road users (VRUs), pose significant challenges for autonomous vehicles in China. VRUs are sometimes unpredictable, taking over different poses and appearing where drivers least expect them. Large animals can suddenly appear on rural roads, while pets may wander onto urban streets. As well as, fully loaded vehicles or tricycles could be difficult to pinpoint the precise vehicle type. Consider the last photo in the center row, it is definitely even very difficult for humans to acknowledge the scene at first sight. The vehicle, loaded with tree branches, is inadvertently in perfect camouflage.
Static road structure and topology can pose a big challenge for autonomous vehicles as well. For instance, the complex intersection shown here highlights the extent of complexities that should be addressed here. While resembling a screenshot from a sci-fi movie, this intersection is, the truth is, an actual place viewable on Google Earth.
If we zoom in, we are going to find an interesting road element which is maybe unique in China, the Left-turn Waiting Area. It’s designed to extend left turn traffic throughput, allowing more cars to undergo the intersection inside one cycle of traffic light. Note the design will not be symmetric, and every direction are designed individually depending on the traffic pattern. And we are able to even find academic papers about it and its effectiveness. Even though it was proposed out of fine intention, it might be really confusing for brand spanking new drivers and the autonomous driving vehicle.
Turning left at an intersection with a waiting area involves a two-step process. Each of them involves different combination of traffic light signals. Here I only showed probably the most common traffic light pattern. The traffic light combination might be more complex, sometimes involving special traffic lights dedicated to waiting areas.
Now we are able to take a deep dive into all of the corner cases of traffic lights. Traffic lights are perhaps the category of objects with the long-tail corner cases. The perception of traffic lights are complicated for 2 different reasons. First now we have to acknowledge the placement and sort and color of the traffic light, then we also must know out of all of the traffic lights we detected, here now we have six, which one our vehicle should concentrate to. To make this decision, it is important to acquire the proper matching between traffic lights and different lanes.
One special style of light is traffic lights designed for buses. We’ve got to acknowledge them accurately for 2 different reasons. Initially, for planning and control of ego vehicle, we want to acknowledge them with a view to accurately ignore them, as they might carry information conflicting with the lights we should always concentrate to and cause confusion for our autonomous vehicles. Yet to predict how a possible bus nearby would maneuver, we want to know its status accurately as well.
Traffic lights designed for buses in China are available in many forms, including LED lights with labels resembling “BRT,” “SRT,” “Bus,” or a single letter “B”. They also can feature specific Chinese characters like “公交” (bus) or “有轨电车” (monorail), and sometimes include icons depicting a cute little bus. Alongside these features, traffic sign modifiers might also be included, making it essential for autonomous vehicles to detect and recognize these features and associate them accurately with the corresponding traffic lights.
jazz music