近期,不少北京市民发现,日常通勤经过的部分路口信号灯运行节奏显著优化:交叉方向无车时,红灯不再固守固定时长,绿灯切换更及时,部分主干路段甚至可达成连续通行多个绿灯。这一直观的出行感受提高,正是北京正在试点落地的AI智能信控平台带来的实际成效。以“路城相融智践首善”为主题的第四届北京智能交通大会近日举办,将公众视线重新聚焦到城市路口红绿灯及其背后的数字路网建设上。
事故预防
从被动接警到主动辨因
传统信号灯大多遵循固定配时逻辑,无论实时路况如何,绿灯时长都不会灵活调整。而在AI智能信控平台的加持下,信号灯具备了动态研判能力,可借助实时数据调整配时策略,达成绿波通行,大幅减少路口空等的无效时间。
近日,千方科技鲲巢·信控智能体正式落地北京。作为全国首个达成全闭环自主运行的城市信控智能体,该智能体在北京朝阳区62个路口上线两个月后运行数据亮眼:主要路口通行延误平均降低15%以上,最高降幅达30%;动态绿波情形下,干线停车次数降低25.2%。
千方科技研究院副院长冉学均在接受北京商报记者采访时表示,过去很长一段时间,行业内的传统检测器只能完成“感知流量变化”的基础动作:平台能识别出当前路段的车流量、平均车速与历史基线存在差异,却很难进一步判断“到底是什么原因导致了异常”。这种“知其然不知其所以然”的感知能力,让很多交通处置动作始终慢半拍,往往要等到路面民警抵达现场,才能最终确认事件类型,再启动相应的疏导思路,中间的时间差很容易让小事件演变成大范围拥堵。
在北京市智能交通协会、赛文交通网联合主办的本届北京智能交通大会上,行业已形成普遍共识:过去30年国内重特大道路交通事故的有效遏制,核心得益于防控措施的重点突出、精准落地。
中国道路交通安全协会会长王长君在会上明确表示,从“智能交通”走向“智慧交通”的过程中,行业首先要厘清的不是工艺有多先进,而是“智慧交通到底要应对什么实际难题”,先锚定真实业务痛点,再挑选匹配的工艺路径,最终才能真正达成交通平台的安全与畅通。
而“致因不准”这个困扰行业多年的痛点,正在被多模态大模型补上关键一环。冉学均介绍,引入大模型能力之后,平台不再只是简单输出“流量异常”的提示,而是可以结合路侧摄像头的画面、雷达的感知数据、过往的事件样本,自动辨别当前的异常究竟来自交通事故、车辆抛锚、非机动车闯入,还是路口人员的临时争执。在精准识别事件类型和严重程度的基础上,交管部门可以直接匹配相应的处置预案,快速调整周边路口的信号配时,联动路面巡逻警力,让响应动作从“事后追赶”变成“事前预判”。
数据复用
从设备铺建到价值释放
如果说事故预防考验的是智慧交通平台“看懂现场”的眼力,那么数据的深度应用,考验的就是整个行业“盘活存量”的能力。过去十年,全国信号机总出货量达到45万台,海量前端设备的铺设,应对了“有没有感知能力”的基础难题;而这些已经装在路上的设备,其数据有没有真正被用起来,正在成为行业必须回答的长期考题。
北京云星宇交通科技股份有限公司机型规划部副总经理远斯在接受北京商报记者采访时提到,过去很长一段时间,国内交通信息化的重心显著向前端倾斜:各地持续投入资源安装摄像头、信号机等各类路侧感知设备,但后续的数据融合、分析、落地应用环节却相对滞后。很多地方的项目做完之后,大量设备产生的数据只存储在本地,不同平台之间的数据互不连通,数据的价值没有被真正挖掘出来。
因而,他建议,新一轮的智慧交通数字化转型,核心思路并不是把旧设备全部推倒重来,而是要在充分兼容原有平台、复用现有感知设备、尽可能控制新增投入的前提下,打通不同来源数据之间的壁垒,挖掘多元数据背后隐藏的交通特性与运行惯性,让沉淀多年的交通数据真正变成可落地、可复制、可验证的治理能力。
顶层外观层面也在为数据的连通铺路。今年10月12日正式印发的《北京市“十五五”时期现代综合交通运输体系发展规划》,为将来五年首都智慧交通的发展作出了平台部署,明确了下一阶段的核心方向:推动跨部门的数据连通,把AI算法下沉到路口、收费站等一线情形,让落地的应用情形形成可持续的运营模式。这也意味着,北京“十五五”期间的智慧交通建设,核心关键词不再是“新增多少设备”,而是“连通、下沉、可运营”。
最新公开数据表明,当下北京市约1.2万个信号灯中,已有7900个达成联网,五环内及城市副中心更是100%联网。
路城同频
从单点优化到跨域协同
如果说打通数据是路城一体的坚实底座,那么跨部门、跨平台的协同,就是撑起整个体系的核心梁柱。北京市智能交通协会会长张世强在本届大会上提到一个直观的行业变化:从第三届北京智能交通大会的“数智筑基”主题,到本届大会的“路城相融”主题,整个行业的关注重心已经从单点工艺的实验室验证,全面走向公路、城市、出行、产业的一体化协同,这个转变也恰好对应了北京智慧交借助去几年的真实发展节奏。
相比工艺落地,跨主体协同推进往往难度更高。它考验的不是算法先进性,而是体制机制的衔接效率。远斯在采访中坦言,体制机制的创新和业务流程的标准化,是当前路城一体建设过程中最大的难点,也是最核心的创新点。机制理顺之后,路城一体建设还需落地差异化策略,针对北京不同区域的路网形态,尤其是常年拥堵的重点路段完成业务适配,最终达成“一域一策、一路一策、一口一策”的精细化治理,而不是用一套通用思路覆盖所有路口。
这种跨部门协同的实际成效,已经在北京多条重点路段上显现出来。今年“五一”假期期间,京哈高速东四环到东五环路段,借助打通路网运行实时监控、事件自动识别、门架情报板动态发布出行信息等多个情形,整个高速干线的通行效能提高了10%以上,路段拥堵时长显著减少,途经市民的出行感受得到了直观改善。
而让路与城达成同频运行,并不意味着把所有交通决策权都交给算法。冉学均介绍,当下实际运行的智慧交通平台会把实时交通状态划分为五个等级。这个等级划分标准完全由交管部门依据路口的平均车速等核心指标制定,无论算法的智能化程度有多高,高等级情形的最终决策权始终牢牢掌握在人的手中,这也让工艺的应用始终保持在安全可控的范围内。
更远的发展图景已经清晰展开。北京市智慧交通发展中心教授级高工杜勇在会上透露,北京智慧交通的总体目标是,到2030年,推动数智工艺全方位、全链条融合交通运输行业,构建起“要素感知全面化、共性基础集约化、治理服务数智化、创新融合产业化、数字空间安全化”的智慧交通发展格局。
“AI等新工艺的落地应用,已经大幅提高了智慧交通平台在路网态势研判、出行趋势预测、信号控制思路动态优化等维度的核心能力。将来,随着政府部门、行业机构与相关企业进一步打通整合分散在不同平台、不同情形下的信息源与数据源,让各类出行相关数据像城市路网一样联网成片、互联互通,整个智能交通平台的能力完备度与运行效能还将迎来显著提高,最终为市民日常出行带来更顺畅、更贴心的服务感受。”北京市智能交通协会秘书长甄爱武说。
北京商报记者 金朝力 实习记者 张继文
图片来源:视觉中国
本文来源:https://www.bbtnews.com.cn/2026/1030/609771.shtml
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
Recently, many Beijing residents have noticed that the operation of traffic signals at certain intersections during their daily commutes has been significantly optimized: when there are no vehicles in intersecting directions, red lights no longer adhere to fixed durations, green light switches are more timely, and some main trunk roads can even achieve consecutive green light passages. This intuitive improvement in travel experience is the actual result brought about by the AI intelligent signal control platform currently being piloted in Beijing. The 4th Beijing Intelligent Transportation Conference, themed "Road-City Integration, Smart Practice of 'Being First in Goodness'," was recently held, refocusing public attention on urban intersection traffic lights and the construction of the digital road network behind them.
Accident Prevention
From Passive Alarm Reception to Active Cause Identification
Most traditional traffic lights follow fixed timing logic; regardless of real-time traffic conditions, green light durations do not adjust flexibly. However, empowered by the AI intelligent signal control platform, traffic lights now possess dynamic analysis capabilities. By utilizing real-time data to adjust timing strategies, they can achieve green wave passage, significantly reducing the invalid time spent waiting idle at intersections.
Recently, the Qianfang Technology "Kunchao" Signal Control Agent was officially launched in Beijing. As the nation's first urban signal control agent to achieve full closed-loop autonomous operation, its operational data has been impressive since going live at 62 intersections in Beijing's Chaoyang District two months ago: average travel delay at major intersections decreased by over 15%, with a maximum reduction of 30%; under dynamic green wave conditions, the number of stops on main lines decreased by 25.2%.
Ran Xuejun, Vice Dean of the Qianfang Technology Research Institute, stated in an interview with the Beijing Business Today that for a long time in the past, traditional detectors in the industry could only perform the basic action of "perceiving traffic flow changes." While the platform could identify that the current road section's traffic volume and average speed differed from historical baselines, it was difficult to further judge "exactly what caused the anomaly." This type of perception capability—knowing the "what" but not the "why"—caused many traffic handling actions to lag behind. Often, it was necessary to wait for road police to arrive at the scene to finally confirm the type of incident before initiating corresponding diversion strategies; this time gap easily allowed minor incidents to evolve into large-scale congestion.
At this Beijing Intelligent Transportation Conference, jointly hosted by the Beijing Intelligent Transportation Association and 7its, a consensus was formed within the industry: the effective containment of major road traffic accidents in China over the past 30 years has been core to the prominent focus and precise implementation of prevention and control measures.
Wang Changjun, President of the China Road Traffic Safety Association, explicitly stated at the meeting that in the process of moving from "Intelligent Transportation" to "Smart Transportation," the first thing the industry must clarify is not how advanced the technology is, but "what actual difficulties Smart Transportation needs to address." Only by first anchoring real business pain points, then selecting matching technological paths, can the safety and smoothness of the transportation platform be truly achieved.
And the pain point of "inaccurate cause determination," which has plagued the industry for years, is being filled by multi-modal large models. Ran Xuejun introduced that after introducing large model capabilities, the platform no longer simply outputs "traffic anomaly" alerts but can combine roadside camera images, radar perception data, and past event samples to automatically distinguish whether the current anomaly comes from a traffic accident, vehicle breakdown, non-motor vehicle intrusion, or a temporary dispute among intersection personnel. Based on accurately identifying the event type and severity, traffic management departments can directly match corresponding handling plans, quickly adjust signal timing at surrounding intersections, and link with road patrol forces, transforming response actions from "after-the-fact pursuit" to "pre-emptive prediction."
Data Reuse
From Equipment Deployment to Value Release
If accident prevention tests the "eyesight" of the smart transportation platform in "understanding the scene," then the deep application of data tests the entire industry's ability to "revitalize existing stock." Over the past decade, the total shipment of traffic signal controllers nationwide reached 450,000 units. The massive deployment of front-end equipment addressed the basic problem of "having perception capabilities." However, whether the data from these devices already installed on the road is truly being utilized is becoming a long-term question that the industry must answer.
Yuan Si, Deputy General Manager of the Model Planning Department at Beijing Yunxingyu Transportation Technology Co., Ltd., mentioned in an interview with the Beijing Business Today that for a long time, the focus of domestic traffic informatization tilted significantly toward the front end. Various regions continuously invested resources in installing cameras, signal controllers, and other roadside sensing equipment, but the subsequent data fusion, analysis, and application implementation stages lagged relatively behind. In many places, after projects were completed, data generated by massive amounts of equipment was only stored locally; data between different platforms was not connected, and the value of the data was not truly excavated.
Therefore, he suggested that the core idea of the new round of digital transformation in smart transportation is not to tear down old equipment and start over, but to break down barriers between data sources under the premise of fully compatibility with original platforms, reusing existing sensing equipment, and controlling new investments as much as possible. By excavating the traffic characteristics and operational inertia hidden behind diverse data, traffic data accumulated over many years can truly be turned into governance capabilities that are implementable, reproducible, and verifiable.
The top-level design is also paving the way for data connectivity. The "Beijing '15th Five-Year Plan' Modern Comprehensive Transportation System Development Plan," officially issued on October 12 this year, has made platform deployments for the development of the capital's smart transportation over the next five years, clarifying the core direction for the next phase: promoting cross-departmental data connectivity, sinking AI algorithms to front-line scenarios such as intersections and toll stations, and allowing implemented application scenarios to form sustainable operational models. This also means that the keywords for Beijing's smart transportation construction during the "15th Five-Year Plan" period are no longer "how much new equipment," but rather "connectivity, sinking, and operability."
The latest public data shows that among Beijing's approximately 12,000 traffic lights, 7,900 have achieved networking. Within the Fifth Ring Road and the City Sub-center, the networking rate is 100%.
Road-City Synchronization
From Single-Point Optimization to Cross-Domain Collaboration
If breaking down data silos is the solid foundation for road-city integration, then cross-departmental and cross-platform collaboration is the core beam and pillar supporting the entire system. Zhang Shiqiang, President of the Beijing Intelligent Transportation Association, mentioned an intuitive industry change at this conference: from the 3rd Beijing Intelligent Transportation Conference's theme of "Digital-Intelligent Foundation Building" to this conference's theme of "Road-City Integration," the industry's focus has shifted from lab verification of single-point technologies to the full integration of highways, cities, travel, and industry. This shift corresponds exactly to the actual development rhythm of Beijing's smart transportation over the past few years.
Compared to technology implementation, cross-entity collaborative advancement is often more difficult. It tests not the advancement of algorithms, but the efficiency of institutional and mechanism linkage. Yuan Si admitted in the interview that the innovation of institutional mechanisms and the standardization of business processes are the biggest difficulties and the core innovation points in the current road-city integrated construction. Once the mechanism is streamlined, road-city integration still requires the implementation of differentiated strategies. It requires business adaptation based on the road network morphology of different regions in Beijing, especially key sections prone to chronic congestion, ultimately achieving the refined governance of "one policy per region, one policy per road, one policy per intersection," rather than using a universal approach to cover all intersections.
The actual effectiveness of this cross-departmental collaboration has already appeared on several key road sections in Beijing. During this year's "May Day" holiday, the section of the Beijing-Harbin Expressway between the East Fourth Ring Road and the East Fifth Ring Road benefited from the integration of road network operation real-time monitoring, automatic event identification, and dynamic release of travel information on gantry information boards. The traffic efficiency of the entire expressway trunk line increased by over 10%, congestion duration on the section was significantly reduced, and the travel experience of passing citizens was intuitively improved.
However, achieving synchronized operation between roads and the city does not mean handing over all traffic decision-making rights to algorithms. Ran Xuejun explained that the currently operating smart transportation platforms divide real-time traffic states into five levels. These level classification standards are entirely formulated by traffic management departments based on core indicators such as average speed at intersections. Regardless of how intelligent the algorithm is, the final decision-making power in high-level scenarios remains firmly in human hands, which ensures that technology application remains within a safe and controllable scope.
A more far-reaching development blueprint has already unfolded clearly. Du Yong, Professor-level Senior Engineer at the Beijing Smart Transportation Development Center, revealed at the meeting that Beijing's overall goal for smart transportation is to promote the comprehensive, full-chain integration of digital-intelligent technology with the transportation industry by 2030, constructing a development pattern characterized by "comprehensive element perception, intensive common foundations, digital-intelligent governance services, industrialized innovation integration, and secure digital space."
"The implementation and application of new technologies like AI have already significantly improved the core capabilities of smart transportation platforms in dimensions such as road network situation analysis, travel trend prediction, and dynamic optimization of signal control strategies. In the future, as government departments, industry institutions, and related enterprises further integrate information and data sources dispersed across different platforms and scenarios—making various travel-related data interconnected like the urban road network—the capability completeness and operational efficiency of the entire intelligent transportation platform will see significant improvement, ultimately bringing smoother and more intimate service experiences to citizens' daily travel." said Zhen Aiwu, Secretary-General of the Beijing Intelligent Transportation Association.
Beijing Business Today Reporter: Jin Chaoli; Intern Reporter: Zhang Jiwen
Image Source: VCG

