您正在使用IE低版浏览器，为了您的雷峰网账号安全和更好的产品体验，强烈建议使用更快更安全的浏览器
此为临时链接，仅用于文章预览，将在时失效
0
导语：目前，卓驭乘用车客户已经达到22家，累计量产车型超过60款，定点车型超过110款。
10月29日晚，智能驾驶方案商卓驭科技正式发布商用车智能辅助驾驶解决方案“卓驭磐行”，并公布了商用重卡、客车、轻卡、ROBOVAN等业务进展。
此次卓驭磐行分为PDS 4和PDS 5两个版本。
PDS 4对应高悟性大模型4.0及以上版本，主要支持高速NOA，面向重卡干线物流。目前这一版本已经实现量产。
PDS 5则对应卓驭下一代原生多模态基础模型，在高速NOA之外，还将支持城区NOA，主要面向客车、轻卡、VAN等需要大量城市道路行驶的商用车型。
卓驭为什么选择在这个时间进入商用车？卓驭科技CEO沈劭劼表示，目前国内商用车智能驾驶渗透率不足1%，而乘用车L2+渗透率已经超过30%。他预计，未来两到三年，商用车智能驾驶会进入更快的增长阶段。
过去几年，乘用车智能驾驶已经经历了一轮快速普及。摄像头、毫米波雷达、域控制器和高算力芯片的成本持续下降，端到端、大模型等技术也已经进入量产车型。
对于已经拥有乘用车客户和量产数据的供应商来说，把已有技术迁移到商用车，比重新开发一套系统成本更低。
在沈劭劼看来，卓驭科技的重要优势是数据。
商用车数据采集成本明显高于乘用车，除了车辆本身，油费和过路费也更高。在业务早期、商用车自身数据体系尚未建立之前，如果完全依靠商用车重新采集数据，投入会非常高。而卓驭科技已有乘用车数据和模型基础，可以降低进入商用车市场的前期成本。
但商用车智能驾驶的难度并不低。
卓驭科技商用车业务副总裁刘思康表示，商用重卡智能驾驶至少有五个难点：模型数据少、感知距离要求高、安全容错率低、车辆控制难度大，同时成本非常敏感。
一辆满载重卡重量可以达到40吨左右，高速行驶时车身长、盲区大、制动距离也更长。此外，重卡空载和满载时的重心和动力学状态差异较大。对于17米左右的半挂车，转弯和变道时还需要处理内轮差、挂车摆动等问题。
卓驭没有针对商用车重新开发一套完全独立的智能驾驶系统，而是直接迁移乘用车已经训练的大模型。
刘思康将其技术路线概括为三句话：底层统一、场景增强、工程适配。
卓驭认为，乘用车和商用车虽然车型不同，但主要都运行在铺装道路，道路结构识别、交通参与者识别、风险判断等基础能力具有较强共性。
目前卓驭乘用车智能驾驶系统已经积累超过30亿公里数据，在此基础上，卓驭再加入约4亿帧重卡特色数据，对重卡干线运输场景进行针对性训练。据悉，这套方案已经实现商用重卡端到端大模型量产。
第二部分是工程适配，由于重卡制动距离更长，对远距离感知能力的要求也更高。
此次磐行方案搭载了卓驭的“激目2.0”系统。相比把激光雷达安装在车外，激目2.0被放置在驾驶舱内，可以减少雨雪、泥水和碎石对传感器的影响。
按照卓驭公布的数据，这套系统在高速场景下的探测距离可以达到500米，并能够识别远处轮胎、纸箱等低反射、小尺寸障碍物。
车辆控制同样需要重新适配。卓驭通过视觉识别挂车姿态，并估算车头和挂车之间的夹角，同时结合实时载重信息，调整车辆转向、制动和动力控制。
针对上下坡、连续弯道和复杂路面，系统还需要协调发动机、缓速器、辅助制动和脚刹等不同执行机构。
不过，相比乘用车用户，一套辅助驾驶系统如果不能降低油耗、减少事故或者降低司机工作强度，很难成为商用车客户长期购买的配置。因此，卓驭还在磐行中加入了节油算法。
根据卓驭介绍，这套算法使用强化学习训练，根据道路坡度、弯道、车辆动力系统以及周围交通状态，计算收油、给油以及使用缓速器和制动系统的时机。
过去两个月，卓驭跟踪了200多辆量产重卡，累计运营里程超过2000万公里。卓驭公布的数据显示，搭载这套算法后，车辆节油率达到4.5%-7.5%。
按照卓驭的测算，以每天行驶800公里计算，一辆重卡每天可以节省约80元燃油成本，一年节省约2万至3万元。
如果再计算降低司机疲劳、安全事故以及保险成本，卓驭预计，一套智能辅助驾驶系统可以在一年左右收回成本。
目前，卓驭已经与徐工汽车、陕汽集团、中国重汽、一汽解放、福田汽车、东风商用车六家国内主要重卡企业达成合作。据卓驭统计，这六家公司合计占国内商用重卡市场超过93%的份额。
其中，陕汽是首家量产客户。今年6月，首款搭载卓驭磐行PDS 4的陕汽德龙X5000正式量产，目前已经有数百辆车运行在韵达快递干线上。
今年9月，徐工发布汉境新能源重卡，搭载卓驭磐行PDS 4的车型预计今年底交付。中国重汽高端重卡汕德卡同样计划今年底量产下线。
海外市场方面，卓驭已经与中国重汽、福田汽车、宇通客车等客户推进海外合作。其中，与中国重汽合作的欧洲项目已经完成部分合规准备，预计明年落地。
卓驭已经完成了第一批车型量产。接下来，这门生意能否继续扩大，还要看节油、可靠性和回本周期这些指标，能否在更多车队的长期运营中得到验证。雷峰网(公众号：雷峰网)雷峰网
雷峰网原创文章，未经授权禁止转载。详情见转载须知。
为了您的账户安全，请验证邮箱
请验证您的邮箱
本文由 AI宝库自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
You are using a low-end IE browser and strongly recommend a faster and safer browser for your Thunder Peak account security and better product experience.
This is a temporary link, only for article preview, which will expire at the time
0
Introduction: At present, there are 22 customers with a cumulative volume of over 60 car types and 110 spot models.
On the evening of 29 October, the Smart Driving Programr, the official publish commercial car smart-assisted driving solution, “Crucing the Rock”, announced business developments such as commercial heavy cards, passenger cars, light cards and ROBOVAN.
This rock is divided into PDF 4 and PDF 5 versions.
PDS 4 corresponds to high-intensity version LLM4.0 and above, which mainly supports high-speed NOAA and is oriented towards heavy-card trunk logistics. At present, this version is already yielding volume.
PDS 5 corresponds to the mastering of the next generation of primary multi-modular infrastructure models, and will support NOAA in urban areas beyond the high-speed NOA, mainly for commercial vehicles requiring large numbers of urban roads, such as buses, light cards and VAN.
Why did Chewie choose to enter a commercial car at this time? According to CEO Shen, the smart driving penetration rate for commercial vehicles in the country is less than 1 per cent, while the L2+ penetration rate for passenger vehicles has exceeded 30 per cent. He expected that, in the next two to three years, the smart driving of commercial vehicles would enter a faster growth phase.
Over the past few years, the use of smart driving by car has undergone a rapid spread. The cost of cameras, millimetre wave radar, domain controllers and high-calculation chips continues to decline, and end-to-end, LLM, etc., have entered the volume-producing model.
For suppliers that already have passenger and volume data, it is less costly to migrate existing technology to commercial vehicles than to redevelop a system.
In Shenzhen ' s view, the important advantage of mastery is data.
Commercial vehicle data collection costs are significantly higher than for passenger vehicles and, in addition to the vehicles themselves, the cost of fuel and road travel is higher. Until such time as the commercial vehicle's own data system is in place, the investment will be very high if the data are recollected exclusively on commercial vehicles. For example, mastery technology already has a basis for passenger vehicle data and models that can reduce the upfront costs of entering the commercial car market.
But it's not so hard to drive a commercial car.
Liu Si-Kang, Vice-President of Commercial Car Operations in Geo-Technology, said that there were at least five difficult points to drive: low model data, high sensory distance requirements, low safety error tolerance, high vehicle control and high cost.
The weight of a full-load truck can be around 40 tons, and the size of the vehicle, the size of the blind area and the range of brakes can be increased at high speed. In addition, there are significant differences in the weight and kinetic state of the load between empty and full. In the case of semi-mounted vehicles of about 17 metres, the issues of internal wheel differentials and the swinging of trailers also need to be addressed in the turnovers and circuits.
rather than redeveloping a fully independent smart driving system for commercial vehicles, LLM, which has already been trained in the vehicle, is directly relocated.
Liu Si-kang summarized its technical course in three sentences: bottom level unification, scene enhancement, and engineering suitability.
It is considered that, although different in size, both passenger and commercial vehicles operate mainly on paved roads, and that there is a strong commonality between basic capacities such as road structure recognition, traffic participant identification and risk assessment.
On the basis of the current accumulation of over 3 billion kilometres of data in the smart-driving system, the choreography has been added to approximately 400 million frames of weight card feature data for targeted training of heavy-card trunk transport scenes. It is known that the programme has already achieved commercial heavy-card-to-end LLM production.
The second component is engineering suitability, and because of the longer brake distance of the recalculation, the requirements for remote sensing are higher.
The rocking programme carries a well-managed “Stream 2.0” system. The impact of rain, snow and rubble on sensors can be reduced by placing laser radars outside the vehicle, with stimulant 2.0 placed in the cockpit.
According to the data published by choreography, the system can be detected at a distance of up to 500 metres under a high-speed scenario and can identify low-reflective, small-size barriers such as remote tyres, paper boxes and so forth.
Vehicle control also requires re-adaptation. The choreography is visualized and the corner between the header and the trailer is estimated, together with real-time load information to adjust vehicle direction, brake and power control.
The system also requires the coordination of different implementing agencies, such as motors, brakes, auxiliary brakes and foot brakes, for up and down slopes, successive bends and complex surfaces.
However, an auxiliary driving system is difficult to become a long-term purchase configuration for commercial car customers if it does not reduce fuel consumption, reduce accidents or reduce the intensity of drivers ' work, compared to passenger passenger vehicles. As a result, choreography was also added to the rock with the oil-saving algorithm.
According to the choreography, the algorithm uses intensive learning training to calculate the timing of the receipt of oil, the provision of oil and the use of slow-trackers and brake systems based on road slopes, bends, vehicle power systems and the state of the surrounding traffic.
Over the past two months, more than 200 weight cards have been tracked, with a cumulative operating mileage of over 20 million kilometres. According to the data published by Chochurch, vehicle fuel savings have reached 4.5-7.5 per cent after carrying the algorithm.
On the basis of choreography, a heavy card could save about $80 per day in fuel costs and some $20,000 to $30,000 per year in terms of 800 kilometres per day.
If the cost of reducing driver fatigue, safety accidents and insurance were to be calculated again, it was expected that a smart-aided driving system would recover costs in about a year.
At present, Zhuo Jiu Jiu has been working with six major domestic heavy-card companies, including Xu Jijia, the Gas Consortium, the Chinese Heavy Air Force, the first steam release, the Fukuda Car and the East Wind Commercial Vehicle. The combined share of these six companies in the domestic commercial heavy-card market is over 93 per cent, according to master management statistics.
Among them, steam is the first mass customer. In June of this year, the first car carrying the steamed Dragon X5,000, which carries the PDS 4 line, now has hundreds of vehicles running on the junction express.
In September this year, the new energy weight card for Xu, publish, is expected to be delivered by the end of this year, with a model for the PDS 4 of the Chosun Rock. China also plans to cut the production line at the bottom of the year.
With regard to overseas markets, Zhuo Jio has advanced overseas cooperation with Chinese customers such as heavy steam, Fukuda and Woo Tong. Among them, the European project for cooperation with China's heavy gas has completed partial compliance preparations and is expected to land next year.
The first car-type production has been completed. Next, whether the business will continue to expand will depend on the indicators of fuel conservation, reliability and return to the current cycle, and will be validated in the long-term operation of more fleets. Thunder Peak Network (WeChat Official Account:Ray Peak Network)
Original Ray Peak article, unauthorized reproduction. For further information, see the information below.
For your account security, please verify the mailbox.
Please verify your mailbox.
Auto-fetched and translated by AI Gem Box. All rights belong to the original author.