— Inside Euro NCAP's 2026 Speed Assistance Rules
In January 2026, Euro NCAP rolled out the biggest overhaul in its history. The star rating is no longer dominated by a single crash performance; it now rests on four pillars — Safe Driving, Crash Avoidance, Crash Protection and Post-Crash Safety — and a car must pass every one of them. Buried inside the Safe Driving stage is an unassuming but remarkably sharp instrument: Speed Assistance, worth 20 points.
It accounts for only a fifth of one stage, yet it may be the single best filter for a carmaker's real technology stack. Here is why these 20 points change the game.
From "fitted" to "accurate"
The EU's General Safety Regulation already mandates Intelligent Speed Assistance on every new car. But regulation only asks whether the system exists. Euro NCAP 2026 asks how good it is — and verifies it in a way that is nearly impossible to game.
The accuracy of the Speed Limit Information Function (SLIF) is worth 4 points and is measured by the test laboratory on public roads: a route defined by Euro NCAP, at least 2,000 km long, crossing three to five countries, with at least 20% of the mileage in each country, at least 10% driven at night, and a prescribed mix of motorway, interurban and urban roads. The test car carries four cameras, RTK-corrected GNSS and a 128-beam LiDAR to build an independent ground truth, against which the speed limit shown on the car's own display is compared frame by frame. Two KPIs, 2 points each: the distance share with the correct limit displayed must exceed 80%, and so must the share of correctly identified speed-limit events.
Consider what this means. Sign designs differ across borders; implicit limits follow national rules; construction-zone gantries vary wildly. A 2,000 km multi-country blind test leaves nowhere to hide for any system tuned only for its home market. For manufacturers targeting Europe — Chinese brands expanding westward in particular — this is a live-fire exam in regional adaptation.
Advanced limits: cameras and maps, neither is optional
One layer deeper, 3 points go to "advanced speed limits": conditional limits for rain, time of day, lane-specific arrows and vehicle categories (2 points), implicit limits inferred from road type — motorway entry, city boundaries, residential zones (0.5) — and dynamic limits on variable message signs (0.5). Manufacturers must demonstrate the system works in 80% of typical driving across ten named countries (Austria, France, Germany, Italy, Luxembourg, the Netherlands, Spain, Sweden, the UK and Norway) plus at least half of the entire Application Area.
Chasing these points exposes the weakness of any single-sensor strategy. A camera can read a "110 when raining" plate but cannot know which rule currently applies; a static map cannot see the roadworks sign erected this morning. The only viable answer is fusion: camera perception, a fresh map, and vehicle state signals working together.
Another 2 points reward data freshness directly: continuously streamed speed-limit data earns 2 points; over-the-air updates at least quarterly earn 1. A car shipped with a buy-once offline map caps out at 1 point here — and will almost certainly bleed points across SLIF accuracy, conditional and dynamic limits too. For the first time, map freshness carries an explicit price tag in the star rating.
V2X leaves the trade show and enters the score sheet
The most interesting 3 points cover Local Hazard warnings. Ten hazard classes — construction zones, objects on the road, broken-down vehicles, post-crash scenes, poor weather, wrong-way drivers and more — and the car must both receive and warn in good time and automatically report what it senses (driver-reported events earn nothing).
Two channels are recognised: cloud exchange, benchmarked against the DFRS (Data for Road Safety) ecosystem — the European data-sharing alliance of carmakers, map providers, telecom operators and road authorities — and direct communication via ITS-G5 or C-V2X with proven C-ROADS interoperability. A single channel is capped at 2.5 points; only cloud and direct communication together unlock the full 3.
V2X has spent fifteen years in pilot corridors and press releases. Euro NCAP just converted it into production-car currency — and pointedly requires an open, shared ecosystem rather than a proprietary cloud. Data silos score zero.
Eight points of speed control, and a hidden speedometer clause
The remaining 8 points reward the Speed Control Function: full marks for iACC — adaptive cruise that automatically adopts the detected limit — versus only 5 for an Intelligent Speed Limiter that isn't on by default. The requirements are concrete: adopt a new limit within 5 seconds, driver offset capped at 10 km/h, kick-down always available to override, an acoustic warning whenever the set speed cannot be held.
The clause most likely to be overlooked: speedometer accuracy must be within −3/+0 km/h, or all SCF points are halved. Speedometers have traditionally over-read by 5–10%, and nobody cared. But once the cruise set-speed comes straight from speed-limit data, an optimistic speedometer silently erodes the real-world benefit. Euro NCAP turned calibration into a 4-point lever — a "minor detail" that can cost a meaningful slice of a star.
What these 20 points actually select for
Lay the points out and the pattern is unmistakable: 12 of 20 hang on SLIF, and SLIF is fundamentally a data capability — sensor fusion, map freshness, cloud coverage, cross-border adaptation. The other 8 test whether a manufacturer commits that data to actual control (default iACC, not an ISL buried in a menu) and to engineering rigour (speedometer calibration). The full-score recipe reads roughly: camera-plus-connected-map SLIF, DFRS and C-V2X on both channels, streamed limit data, iACC as standard, a −3/+0 speedometer. Few vehicles on sale today assemble every piece.
Zoom out, and Speed Assistance is the 2026 philosophy in miniature. Safety is no longer defined by steel and airbags at the moment of impact, but by a data loop running the entire journey: the car knows the rules of the road, knows what is happening on it, and gently but firmly helps the driver stay inside the envelope. Speeding remains one of Europe's leading killers; behind these 20 points is the reclassification of "knowing the limit" from a convenience feature into a safety function.
For carmakers, the room for regulatory box-ticking is vanishing. The 2026 soft-landing (60% suffices for five stars in Safe Driving) is the last buffer; full thresholds bite from 2028. The data-infrastructure bill has already started accruing.
Based on the Euro NCAP Safe Driving – Vehicle Assistance Protocol v1.1 (October 2025), Technical Bulletins SD-303 and G001, and public DFRS materials; the protocols take effect January 2026. Views are my own.
——解读 Euro NCAP 2026 速度辅助新规
2026年1月,Euro NCAP 史上最大规模的评价体系改革正式落地。星级不再由一次碰撞成绩主导,而是拆成安全驾驶、碰撞避免、碰撞保护、碰撞后安全四门"必修课",每门都不能挂科。在这套新体系里,有一个不起眼却极具杀伤力的科目:速度辅助(Speed Assistance),20分。
它只占安全驾驶阶段的五分之一,却可能是最能筛选车企真实技术底盘的一项。这篇文章把这20分拆开,讲讲它为什么会改变游戏规则。
从"装了就行"到"准了才算"
欧盟 GSR II 法规早已强制所有新车配备 ISA(智能速度辅助),但法规只管"有没有"。Euro NCAP 2026 管的是"好不好"——而且用了一种几乎无法作弊的方式。
限速识别(SLIF)的精度分值4分,由实验室在公开道路实测:路线由 Euro NCAP 指定,总长至少 2000 公里,横跨3到5个国家,每个国家至少占20%里程,至少10%在夜间,高速、城际、市区按比例分配。测试车加装四路摄像头、RTK 高精度定位和128线激光雷达,独立建立"真值",逐帧比对车机显示的限速与真实限速。两个指标各值2分:显示正确限速的里程占比要超过80%,正确识别的限速事件占比也要超过80%。
这意味着什么?各国限速牌样式不同、隐含限速规则不同、施工区电子牌五花八门——2000公里的跨国盲测,让任何"只在本国调优"的系统无处遁形。对志在欧洲市场的车企(尤其是出海的中国品牌),这是一场真刀真枪的区域适配考试。
高级限速:摄像头和地图,一个都不能少
再往深一层,3分给了"高级限速":雨天限速、时段限速、分车道箭头限速、按车型区分的限速(条件限速,2分),无牌路段靠道路类型推断的隐含限速(0.5分),以及可变情报板的动态限速(0.5分)。厂商要证明系统在奥、法、德、意、卢、荷、西、瑞典、英、挪十国,外加应用区一半以上国家,80%的典型工况下有效。
想拿这3分,纯技术路线的短板暴露无遗。纯摄像头读得出"雨天110"的牌子,却不知道现在有没有下雨、哪条规则生效;纯地图数据跟不上今天刚立起来的施工限速牌。唯一的解是融合:摄像头感知 + 高鲜度地图 + 车辆状态信号三方打通。
还有2分给"系统更新":行驶中持续流式获取限速数据得2分,至少每季度OTA更新得1分。换句话说,买断式离线地图的车型,在这条上天花板就是1分——而且它的 SLIF 精度、条件限速、动态限速大概率也会全面失血。地图的"鲜度"第一次被明码标价写进了星级。
V2X 走出展台,走进评分表
最有趣的是"本地危险警告"3分。施工区、路面异物、抛锚车、事故现场、恶劣天气、逆行车等十类危险,车辆要能接收并提前提示,还要能自动上报(驾驶员手动标记不算分)。
通道有两条:云端参照 DFRS(Data for Road Safety)生态——由车厂、图商、电信运营商和欧洲公共部门组成的安全数据共享联盟;直连则是 ITS-G5 或 C-V2X,须证明与 C-ROADS 路侧设施互通。单通道封顶2.5分,云端+直连双通道才能拿满3分。
V2X 讲了十几年,大多停留在示范区和发布会。Euro NCAP 用3分把它变成了量产车的真金白银,而且明确要求接入开放共享生态而非厂商私有云——数据孤岛不给分。
8分的速度控制,和一条隐藏的"表显条款"
剩下8分给速度控制(SCF):iACC(自适应巡航自动跟随限速)满分8分,非默认开启的智能限速器 ISL 只有5分。要求很具体:限速变化后5秒内采纳、驾驶员偏置不超过10 km/h、深踩油门可随时超越、无法维持限速时必须声音警告。
最容易被忽略的是这一条:速度表精度必须达到 -3/+0 km/h,否则 SCF 分数减半。长期以来车速表普遍"虚高"5-10%,这在传统评价里无关痛痒。但当巡航速度直接由限速数据设定时,表显虚高意味着实际控速效益打折——Euro NCAP 干脆把标定精度变成了4分的杠杆。一个"小事"砍掉一颗星的零头,就是这么来的。
这20分在筛选什么
把分数摊开看,答案很清楚:12分系于 SLIF,而 SLIF 的本质是数据能力——感知融合、地图鲜度、云服务覆盖、跨国适配。剩下8分考的是把数据用于控制的诚意(默认iACC而非藏在菜单里的ISL)和工程严谨(表显标定)。满分路径大致是:摄像头+联网地图融合的SLIF、接入DFRS与C-V2X双通道、流式限速更新、默认iACC、-3/+0的速度表。今天能凑齐全部拼图的车型并不多。
更宏观地看,速度辅助是2026新体系设计哲学的缩影:安全不再是碰撞瞬间的钢板与气囊,而是行程全程的数据闭环——车知道路的规则,知道路上正在发生什么,并温和而坚定地帮驾驶员守住边界。超速仍然是欧洲道路死亡的头号成因之一,这20分背后,是把"知道限速"从体验功能重新定义为安全功能。
对车企来说,留给"应付法规"的空间正在消失。2026年软着陆期(安全驾驶5星门槛暂降至60%)是最后的缓冲,2028年起全额执行。数据基础设施的账,现在就得开始还了。
本文基于 Euro NCAP Safe Driving – Vehicle Assistance Protocol v1.1(2025年10月)、TB SD-303、TB G001 及 DFRS 公开资料,协议2026年1月实施。观点仅代表个人。