Dickmanns Vision System Steers Mercedes Sedans in Live Paris Traffic

Kaspars Foigts

Real-time 25 Hz computer vision enabled fully automated highway lane changes in unscripted public traffic as early as 1994.

AI-assisted summary of Dickmann's autonomous cars, 1980s — original source. Check the original for full context.

Source published: 2016-01-12

Key ideas

Ernst Dickmanns' VaMoR platform demonstrated autonomous highway driving as early as 1987, reaching speeds of 96 km/h across more than 20 km on an unopened section of the German Autobahn. (0:01)
VaMoR's real-time closed-loop architecture directly operated vehicle actuators—steering, throttle, and braking—while dynamically controlling the gaze direction of onboard vision cameras. (0:01)
Under Phase 3 of the pan-European Eureka PROMETHEUS project (1993–1994), hardware miniaturization allowed full autonomous vision systems to fit inside Mercedes-Benz S-Class passenger sedans. (1:17)
At a 1994 showcase near Paris, the VaMP vehicle tracked up to six surrounding vehicles at 25 Hz to autonomously execute lane-change decisions in unscripted, live three-lane freeway traffic. (1:17)

Detailed notes

## Background Published on YouTube by Kaspars Foigts on January 12, 2016, this video documents the pioneering work of Ernst Dickmanns with the VaMoR and VaMP autonomous vehicle projects. The channel's description notes that it highlights how these computer-guided vehicles managed to navigate thousands of miles autonomously on public roads starting in the 1980s.

## TL;DR The video chronicles the development of early autonomous driving technology under Ernst Dickmanns, beginning with the VaMoR platform which achieved full closed-loop computer control of steering, throttle, and braking by 1986 after eight years of simulation work. In 1987, VaMoR demonstrated sustained autonomous driving at speeds up to 96 km/h over 20 kilometers on a newly constructed German Autobahn, eventually incorporating obstacle tracking and lane-change maneuvers. The program then highlights the Eureka PROMETHEUS project in 1993–1994, where the miniaturized VaMP vision system tracked multiple surrounding vehicles at 25 Hz and executed fully autonomous lane changes in live freeway traffic near Paris.

## Atmosphere & style The presentation carries a calm, matter-of-fact archival documentary tone characteristic of classical European research overviews. The narration delivers technical milestones with clinical precision, avoiding sensationalism while letting the historic footage speak for itself. Watching early computer racks and micro-miniaturized cameras guide full-sized Mercedes sedans through live 1990s highway traffic creates an eerie, retro-futuristic atmosphere that underscores just how far ahead of its time the engineering was.

## Bulletin - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) Ernst Dickmanns' VaMoR platform demonstrated autonomous highway driving as early as 1987, reaching speeds of 96 km/h across more than 20 km on an unopened section of the German Autobahn. - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) VaMoR's real-time closed-loop architecture directly operated vehicle actuators—steering, throttle, and braking—while dynamically controlling the gaze direction of onboard vision cameras. - [01:17](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=77s) Under Phase 3 of the pan-European Eureka PROMETHEUS project (1993–1994), hardware miniaturization allowed full autonomous vision systems to fit inside Mercedes-Benz S-Class passenger sedans. - [01:17](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=77s) At a 1994 showcase near Paris, the VaMP vehicle tracked up to six surrounding vehicles at 25 Hz to autonomously execute lane-change decisions in unscripted, live three-lane freeway traffic.

## Key takeaways - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) Ernst Dickmanns and his research team pioneered early autonomous driving using computer-guided road vehicles in the 1980s. - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) The VaMoR vehicle entered active road testing in 1986 following eight years of foundational simulation development. - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) The vehicle's computer system directly controlled vehicle actuators—steering wheel, throttle, and brakes—via real-time sensor data fusion. - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) VaMoR featured an active vision platform capable of dynamically panning and altering camera gaze under computer control. - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) During 1987 Autobahn trials, VaMoR drove autonomously for over 20 kilometers at speeds up to 96 km/h. - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) The 96 km/h speed threshold in 1987 was limited strictly by engine gearing rather than onboard computational limits. - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) Automated obstacle recognition was introduced in 1989, enabling convoy-following behind preceding vehicles. - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) System capabilities matured to include fully autonomous lane changes by the early 1990s. - [01:17](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=77s) Between 1993 and 1994, research joined Phase 3 of Eureka PROMETHEUS, equipping three Mercedes-Benz S-Class sedans with UBM vision architectures. - [01:17](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=77s) Micro-miniaturized cameras and denser electronics made it possible to house the entire computational payload within a standard vehicle trunk. - [01:17](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=77s) Engineers placed part of the hardware on the right rear seat of the VaMP test car to facilitate bench maintenance during trials. - [01:17](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=77s) VaMP's upgraded vision system tracked up to six surrounding vehicles across both front and rear hemispheres at 25 Hz. - [01:17](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=77s) In October 1994, VaMP demonstrated autonomous lane-change decision-making in normal, unscripted traffic on a three-lane freeway near Paris.

## Detailed walkthrough ### [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) Early Breakthroughs with VaMoR on the Autobahn

The development of automated driving achieved its earliest and most remarkable milestones through computer-guided road vehicles, led by Ernst Dickmanns and his team. The primary experimental testbed, the VaMoR vehicle, entered active field operation in 1986. This rollout was the culmination of eight years of foundational preparatory engineering centered on multiple tiers of simulation, which ultimately converged into a closed, real-time simulation loop. Operating on data fusion gathered from an array of onboard sensors, the vehicle's computer system exercised direct, automated control over the primary driving actuators: the steering wheel, the throttle, and the braking system. In addition to vehicle motion, the platform was capable of active vision, using computer control to dynamically adjust the pan and gaze direction of its onboard cameras.

```mermaid flowchart TD accTitle: VaMoR Actuation and Sensor Loop accDescr: Diagram illustrating sensor data fusion driving actuators and camera direction. subgraph SENSING["Perception & Vision"] A["Active TV Cameras"] B["Vehicle Sensors"] end subgraph PROCESSING["Onboard Computer System"] C["Real-Time Data Fusion Loop"] end subgraph ACTUATION["Vehicle Controls"] D["Steering Wheel"] E["Gas Pedal"] F["Brakes"] G["Camera Direction Control"] end A --> C B --> C C --> D C --> E C --> F C --> G G -.->|"adjusts gaze"| A ```

In 1987, VaMoR set an unprecedented benchmark for automated transit during tests on a newly completed section of the German Autobahn before it was officially opened to public transit. The vehicle navigated autonomously across a span exceeding 20 kilometers, executing maneuvers at varying speeds up to a ceiling of 96 kilometers per hour, constrained solely by the vehicle's engine gearing rather than computer processing bottlenecks. Building on this dynamic highway control, the engineering team expanded the system's scene understanding in 1989 to include automated obstacle recognition, which enabled the platform to safely transition into a convoy-following mode behind leading vehicles. By the dawn of the 1990s, the system's behavioral repertoire had matured further to perform fully autonomous lane changes.

- As the narrator highlights regarding this milestone era: “In 1987 VaMoR demonstrated worldwide top performance on a new stretch of Autobahn not yet turned over to the public.” - Preparatory engineering: 8 years of simulation work leading up to operational deployment in 1986. - 1987 test speed: Cruised at speeds reaching up to 96 km/h over stretches longer than 20 km. - Functional milestones: Automated obstacle detection introduced in 1989; autonomous lane changing realized in the early 1990s.

---

[01:17](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=77s) Project PROMETHEUS and the VaMP Demonstration in Paris

Between 1993 and 1994, the research transitioned into the third phase of the pan-European Eureka PROMETHEUS project, outfitting three Mercedes-Benz S-Class passenger sedans with custom vision architectures engineered by the Universität der Bundeswehr München (UBM). Owing to the development of micro-miniaturized television cameras and denser computing hardware, the entire processing and electronic payload could now be packaged inside a standard vehicle trunk. However, inside the specific UBM test vehicle, named VaMP, engineers elected to install a portion of the hardware package across the right rear passenger seat to maximize component accessibility and speed up bench maintenance during experimental field trials.

```mermaid flowchart LR accTitle: VaMP Traffic Tracking and Decision Architecture accDescr: Flow from multi-camera tracking to autonomous lane change decision. subgraph PERCEPTION["Vision Tracking (25 Hz)"] H["Front Hemisphere Cameras"] I["Rear Hemisphere Cameras"] J["Track up to 6 Vehicles"] end subgraph REASONING["UBM Computer Science Module"] K["Surrounding Traffic State Estimation"] L["Target Neighbor Lane Assessment"] M["Autonomous Lane Change Decision"] end H --> J I --> J J --> K K --> L L --> M ```

The upgraded vision system was capable of tracking up to six external vehicles simultaneously across both the front and rear hemispheres. The relative spatial states, velocities, and trajectories of these surrounding objects with respect to the ego-vehicle were concurrently computed at an update rate of 25 Hz. This high-frequency environmental awareness culminated in a major trial held near Paris in October 1994 during the final PROMETHEUS showcase. Merging into active, unscripted traffic along a three-lane freeway, the vehicle used a specialized software module developed by UBM computer science researchers to monitor traffic across all adjacent lanes and evaluate safe opening gaps, successfully executing fully automated lane-change decisions in live highway conditions.

- As the narrator recounts regarding the milestone trial: “On this basis at the final demonstration of the Prometheus project near Paris in October of 1994 on a three-lane freeway with normal traffic the capability of making the decision for lane change fully autonomously was shown for the first time.” - Fleet integration: Three Mercedes-Benz S-Class sedans integrated with UBM vision hardware during Phase 3 of Eureka PROMETHEUS (1993–1994). - Multi-target tracking: Concurrently tracks up to 6 surrounding vehicles in front and rear hemispheres at 25 Hz. - Public road validation: October 1994 Paris demonstration evaluated automated decision-making and lane changes within live, three-lane freeway traffic.

## Mentioned Ernst Dickmanns, Eureka PROMETHEUS project, German Autobahn, Kaspars Foigts, Mercedes-Benz S-Class, Universität der Bundeswehr München, VaMoR, VaMP

## Video description Short video about Ernst Dickmanns VaMoR and VaMP projects - fully autonomous vehicles, which travelled thousands of miles autonomously on public roads in 1980s.

[Source](https://www.youtube.com/watch?v=_HbVWm7wdmE)

中文简报

迪克曼斯视觉系统在巴黎真实路况中引导梅赛德斯轿车行驶早在1994年，实时25赫兹计算机视觉技术便已能在未预设的公共交通环境中实现全自动高速公路变道。该视频记录了恩斯特·迪克曼斯领导下的早期自动驾驶技术发展历程，始于VaMoR平台。经过八年的仿真研究，该平台于1986年实现了计算机对转向、油门和制动的全闭环控制。1987年，VaMoR在一段新建的德国高速公路上，以高达96公里/小时的速度持续自主行驶了20多公里，随后又加入了障碍物追踪和变道操作功能。项目接着聚焦于1993年至1994年的尤里卡“普罗米修斯”（Eureka PROMETHEUS）项目，在该项目中，小型化的VaMP视觉系统以25赫兹的频率追踪周围的多辆汽车，并在巴黎附近的真实高速公路车流中执行了全自动变道。早在1987年，恩斯特·迪克曼斯的VaMoR平台就展示了高速公路自动驾驶能力，在一段尚未开放的德国高速公路上以96公里/小时的速度行驶了20多公里。 (0:01) VaMoR的实时闭环架构可直接操作车辆执行器（转向、油门和制动），同时动态控制车载视觉摄像头的视线方向。 (0:01) 在泛欧尤里卡“普罗米修斯”项目第三阶段（1993–1994年）中，硬件的小型化使得整套自动驾驶视觉系统得以安装在梅赛德斯-奔驰S级乘用轿车内。 (1:17) 在1994年巴黎附近的一场展示中，VaMP车辆以25赫兹的频率追踪多达六辆周围车辆，在未经预设的真实三车道高速公路车流中自主做出变道决策。 (1:17)

Reading summary

Pioneered by Ernst Dickmanns and his team at the Universität der Bundeswehr München, early autonomous driving research established that real-time computer vision and closed-loop control could guide vehicles at highway speeds and execute complex maneuvers in live traffic. Closed-Loop Control and High-Speed Autobahn Testing (0:01) Field testing began in 1986 with the VaMoR experimental testbed following eight years of foundational simulation work. The vehicle integrated real-time sensor data fusion to directly actuate the steering wheel, throttle, and braking systems, while also employing an active vision mechanism that dynamically adjusted the gaze and pan of onboard cameras. In 1987, VaMoR demonstrated high-speed transit on an unopened section of the German Autobahn, covering more than 20 kilometers at speeds up to 96 kilometers per hour. According to project records, this speed ceiling was imposed by the vehicle's engine gearing rather than limits in computer processing power. Over the subsequent years, the engineering team expanded VaMoR's environmental interpretation capabilities. In 1989, the platform incorporated automated obstacle recognition, allowing the system to operate in a convoy-following mode behind lead vehicles. By the early 1990s, the vehicle's control architecture had evolved to execute autonomous lane-change maneuvers. Multi-Target Tracking and Live Traffic Decisions (1:17) Between 1993 and 1994, the research expanded into the third phase of the pan-European Eureka PROMETHEUS project, deploying custom vision systems across three Mercedes-Benz S-Class sedans. Advances in micro-miniaturized cameras and computing density enabled the entire hardware payload to fit within a standard trunk, though the dedicated VaMP test vehicle retained hardware across the right rear passenger seat to allow accessible bench maintenance during trials. VaMP's upgraded perception system concurrently tracked up to six surrounding vehicles across both the front and rear hemispheres at an update rate of 25 Hz, computing relative positions, velocities, and trajectories. This capability culminated in an October 1994 public trial near Paris, where the vehicle merged into unscripted, normal traffic on a three-lane freeway and successfully made fully autonomous lane-change decisions based on real-time traffic monitoring and gap evaluation. The trials proved that dynamic machine vision could handle high-speed transit and unscripted highway lane changes, though early platforms like VaMP still relied on visible cabin modifications to facilitate maintenance and component access during active field experimentation.

精炼阅读摘要

20世纪80年代至90年代初，Ernst Dickmanns团队通过VaMoR和VaMP试验平台开创了计算机视觉引导的自动驾驶技术，证明了基于实时主动视觉与传感器融合的闭环架构能够在高速公路上实现高速巡航与真实交通流中的自主变道决策。 VaMoR闭环控制与动态巡航突破 (0:01) 在经历长达八年的仿真系统研发后，Dickmanns团队于1986年将VaMoR试验车投入实际运行。该平台通过车载传感器的数据融合构建了闭环实时控制回路，计算机可直接操纵方向盘、油门与制动器，并能主动调节车载摄像机的转动与注视方向。1987年，该车在尚未交付公开通行的德国高速公路（Autobahn）上完成了超过20公里的自主行驶，巡航时速达到96公里；记录显示，这一车速上限受制于车辆引擎传动比，而非计算处理能力的瓶颈。在实现早期高速公路动态控制的基础上，团队于1989年为系统引入自动化障碍物识别能力，使车辆能够进入跟随前车的车队行驶模式。到20世纪90年代初，该系统的行为能力进一步拓展，掌握了全自主变道操作。多目标感知与真实路况自主变道 (1:17) 1993年至1994年，该研究推进至泛欧Eureka PROMETHEUS项目的第三阶段，慕尼黑联邦国防军大学（UBM）为三辆梅赛德斯-奔驰S级轿车装备了定制视觉架构。微型化摄像机与更高密度的计算硬件使得整套电气设备能够装入常规车辆后备箱，不过在UBM专用的VaMP测试车内，部分硬件被布置在右后座以便于试验期间的维护与检修。升级后的视觉系统支持前后半球以25赫兹的高频同时追踪最多6辆外围车辆的相对位置、速度与运动轨迹。1994年10月在巴黎附近三车道高速公路上展开的公开演示中，VaMP利用UBM研发的计算机科学模块，在未经预设脚本的日常交通流中实时监测各车道并评估安全间距，首次公开展示了在真实公路环境中完全自主做出并执行变道决策的能力。 Dickmanns团队的技术路径确立了现代自动驾驶利用多目标视觉追踪与状态估计进行实时规划的基础，而系统对专用高速公路环境的依赖以及早期测试中对车内空间与硬件维护的工程折衷，也展现了早期自主移动系统从实验室走向全面实用化过程中面临的软硬件约束。

Detailed notes — zh

## Background 该视频由 Kaspars Foigts 于 2016 年 1 月 12 日发布在 YouTube 平台上。页面简介说明，这是一部关于恩斯特·迪克曼斯（Ernst Dickmanns）的 VaMoRs 和 VaMP 项目的简短视频，展示了这些完全自动驾驶的车辆如何在 20 世纪 80 年代的公共道路上自主行驶了数千英里。 ## TL;DR 视频回顾了 20 世纪 80 至 90 年代计算机视觉与自动驾驶先驱恩斯特·迪克曼斯团队的开创性成就。1986 年投入实测的 VaMoRs 试验车实现了油门、刹车、转向及摄像机朝向的电控闭环接管，并在 1987 年未开放的高速公路上以高达 96 km/h 的速度完成了超过 20 公里的全自动巡航。随着 1990 年代尤里卡计划“Prometheus”项目的推进，研发团队在奔驰 S 级改装车 VaMP 上实现了硬件微型化，并借助 25 Hz 刷新率的高频双向多目标追踪系统，于 1994 年巴黎真实开放高速公路车流中成功完成了世界首次完全自主变道演示。 ## Atmosphere & style 全片洋溢着浓厚且严谨的历史工程纪实感，旁白沉稳、客观而专业，充满学术纪录片的严密与克制。画面记录着早期笨重的显像管监视器、精密机械接管的方向盘以及奔驰原型车在车流中平稳穿行的珍贵镜头，将冷酷的技术参数与极具时代感的实验影像融为一体。整部视频散发着一种不动声色的历史震撼力，让观者深切感受到几十年前科研人员在自动驾驶黎明期所展现出的超前远见与突破性野心。 ## Bulletin - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) 恩斯特·迪克曼斯团队的 VaMoRs 试验车在 1987 年的德国高速公路上，以高达 96 km/h 的速度完成了超过 20 公里的全自动无干预连续行驶里程碑。 - [01:10](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=70s) 1994 年 10 月，在欧洲尤里卡“Prometheus”项目下，VaMP 试验车在巴黎真实三车道高速公路交通流中首次向全球展示了完全自主变道决策与执行。 - [01:10](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=70s) 慕尼黑联邦国防军大学研发的双向视觉追踪系统具备极高计算效率，能以 25 Hz 刷新率在前向与后向视野中同时并发追踪多达 6 辆周边车辆的相对运动状态。 - [01:10](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=70s) 硬件微型化技术在 1990 年代初取得重大进展，原本庞大的机柜级电子设备得以全部整合入梅赛德斯-奔驰 S 级轿车的后备箱内。 ## Key takeaways - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) 自动驾驶道路车辆制导的核心工程突破早在 20 世纪 80 年代即已启动并初见成效。 - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) VaMoRs 测试车在 1986 年正式路测前，经历了长达 8 年的仿真系统与实时闭环技术研发积累。 - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) 系统实现了对车辆底盘控制的全面软硬件闭环接管，涵盖方向盘转向、油门开度、刹车制动以及摄像头朝向调节。 - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) 1987 年在未通车的德国高速公路测试中，VaMoRs 的最高行驶车速（96 km/h）受限于发动机转速而非自动驾驶算法。 - [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) 1989 年算法升级引入了障碍物识别能力，促成了系统向车队跟随巡航模式的顺利演进。 - [01:10](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=70s) 1993 至 1994 年间，团队借助欧洲尤里卡计划的 Prometheus 项目，将视觉自动驾驶系统装载于梅赛德斯-奔驰 S 级轿车上展开工程验证。 - [01:10](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=70s) 微型化电视摄像机的使用大幅缩减了设备体积，使复杂的计算机与感知硬件能够完整收纳进普通轿车后备箱。 - [01:10](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=70s) 在 VaMP 改装车上，研发团队将部分设备外置安装于后排右侧座位，以提升路测调试的可维护性与操作便利度。 - [01:10](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=70s) 视觉感知模块具备双向半球追踪能力，前后视野各支持追踪多达 6 辆目标车辆并计算其相对运动态势。 - [01:10](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=70s) 系统的状态估算刷新率达到 25 Hz，确保了对高速复杂动态路况的超高实时响应能力。 - [01:10](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=70s) 1994 年在巴黎高速公路的演示标志着自动驾驶汽车首次在包含普通社会车流的三车道环境下成功自主完成变道。 - [01:10](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=70s) 由慕尼黑联邦国防军大学（UBM）开发的核心计算模块，通过全天候监控自车周边及相邻目标车道车流，确保了自主切入变道的安全性。 ## Detailed walkthrough ### [00:01](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=1s) VaMoRs 试验车与 1980 年代的高速公路突破在自动化道路车辆制导领域，最引人注目的突破始于 20 世纪 80 年代。旁白开篇强调：“The most striking successes have been achieved in autonomous road vehicle guidance.” 这一系列开创性成果的核心测试平台是 VaMoRs 试验车，它自 1986 年起正式投入实际运行。在此之前，研发团队进行了长达 8 年的前期技术积累，探索了各种类型的模拟系统，最终集成为闭环的实时仿真环境。在硬件与控制逻辑上，该车依赖多传感器的数据融合（Data Fusion），车载计算机系统能够全自动接管车辆的核心操控，直接控制方向盘转向、油门踏板开度以及制动系统。不仅如此，摄像机的拍摄朝向也可以完全由车载计算机根据行驶感知需要自动控制调节。进入 1987 年，VaMoRs 在一段尚未向公众开放交付的全新德国高速公路（Autobahn）路段上，向世界展示了当时全球顶尖的自动驾驶性能。试验车以多种不同车速进行了巡航测试，受限于当时的发动机最高转速，其最高行驶时速达到了 96 km/h，在完全没有人类干预的情况下连续行驶了超过 20 公里。随着感知技术的迭代，系统从 1989 年起获得了识别道路障碍物的能力，并顺理成章地实现了向车队跟随模式（Convoy driving mode）的平滑过渡。 - **前期技术积累**：在 1986 年 VaMoRs 实际公路路测前，经历了长达 8 年的各类模拟与实时仿真闭环研发。 - **底层控制集成**：车载系统对油门、刹车、方向盘以及摄像头朝向实现了全电控闭环控制。 - **1987 年高速里程碑**：在未通车的 Autobahn 上实现超过 20 公里的全自动巡航，最高时速达 96 km/h（受发动机转速限制）。 - **功能迭代演进**：1989 年突破障碍物检测，随后解锁了自动巡航跟随车队行驶的能力。 --- ### [01:10](https://www.youtube.com/watch?v=_HbVWm7wdmE&t=70s) Eureka Prometheus 项目与 VaMP 的自主变道突破到了 1990 年代初期，研发团队进一步攻克了车辆完全自主变道的关键难题。在 1993 年至 1994 年期间，作为欧洲尤里卡计划（Eureka）框架下“Prometheus”项目第三阶段的一部分，团队将慕尼黑联邦国防军大学（Universität der Bundeswehr München, UBM）开发的视觉系统装载在三辆梅赛德斯-奔驰 S 级（Mercedes-Benz S-Class）轿车上。得益于新型微型化电视摄像机（Micro-miniaturized TV cameras）的应用，原本庞大的电子计算设备如今已经可以全部容纳在轿车后备箱中。不过，在 UBM 专门改装的试验车 VaMP 上，出于调试可维护性与易操作性的考虑，部分电子设备被额外安装在了后排右侧座椅位置。这套先进的双向多目标视觉追踪系统展现出了惊人的感知处理能力。旁白指出：“Up to six other vehicles may be tracked in parallel in both the front and the rear hemisphere.” 无论是在车辆前向半球还是后向半球的视野内，系统都能同时追踪多达 6 辆其他车辆，并以 25 Hz 的超高刷新率并发估算目标车辆相对于本车（Ego vehicle）的相对运动状态。1994 年 10 月，在巴黎附近一条拥有正常社会车流的三车道高速公路上，Prometheus 项目迎来了最终演示，团队向全世界首次公开验证了自动驾驶车辆在开放交通流中全自动完成变道决策与执行的能力。由 UBM 计算机科学团队专门构建的核心模块负责全天候实时监视自车周边的整体交通态势，尤其能够高精度掌握目标相邻车道的车流动态，从而安全平顺地完成切入。 ```mermaid flowchart TD accTitle: VaMP 自动变道系统感知与决策架构 accDescr: 展示环境感知输入通过 25Hz 双向视觉追踪，最终生成自主变道控制决策的流程 subgraph Perception ["环境感知输入"] Cams["微型化电视摄像机"] Sensors["车载辅助传感器"] end subgraph Tracking ["UBM 视觉追踪与状态估计"] Fusion["25 Hz 并发解算核心"] FrontTrack["前向视野追踪 (≤6辆)"] RearTrack["后向视野追踪 (≤6辆)"] TargetLane["相邻变道车道监控"] end subgraph Execution ["决策与底盘控制"] Decision["完全自主变道决策"] Actuators["执行机构: 转向 / 油门 / 制动"] end Cams --> Fusion Sensors --> Fusion Fusion --> FrontTrack Fusion --> RearTrack Fusion --> TargetLane FrontTrack --> Decision RearTrack --> Decision TargetLane --> Decision Decision --> Actuators ``` - **Prometheus 项目合作**：1993–1994 年期间，基于奔驰 S 级轿车搭载 UBM 视觉系统进行深入工程化改造。 - **硬件小型化**：微型化摄像机使原本机柜级的电子系统得以浓缩塞入轿车后备箱，VaMP 车内则分设于后座右侧以便实时维护。 - **高频多目标追踪**：前后视野可并发锁定多达 6 辆移动目标车辆，状态估算频率稳定在 25 Hz。 - **1994 年巴黎实测**：在真实三车道繁忙高速路况下，首次展示了车辆基于周边与相邻车道车况自主做出的自动变道决策。 ## Mentioned Ernst Dickmanns, VaMoRs, VaMP, Eureka Prometheus, 慕尼黑联邦国防军大学 (Universität der Bundeswehr München), 梅赛德斯-奔驰 S 级 (Mercedes-Benz S-Class), Autobahn ## Video description Short video about Ernst Dickmanns VaMoR and VaMP projects - fully autonomous vehicles, which travelled thousands of miles autonomously on public roads in 1980s. [Source](https://www.youtube.com/watch?v=_HbVWm7wdmE)

Original transcript

[0:01] the most striking successes have been [0:02] achieved in autonomous road vehicle [0:05] Guidance the test vehicle vors seen here [0:09] has been in operation since [0:11] 1986 this period was preceded by 8 years [0:14] of Preparatory developments with [0:16] different types of simulation [0:18] culminating in the realtime simulation [0:21] Loop Based on data Fusion from various [0:24] sensors the onboard computer system [0:27] automatically controls the steering [0:29] wheel the gas pedal and the [0:31] brakes the direction the cameras are [0:34] pointing may also be controlled by [0:36] computers in 1987 vomor demonstrated [0:39] worldwide top performance on a new [0:42] stretch of Autobon not yet turned over [0:44] to the public it cruised at various [0:47] speeds with top speed of 96 kmph limited [0:51] only by engine speeds over a distance of [0:54] more than 20 [0:56] km since 1989 it has also been possible [0:59] I to recognize [1:02] obstacles as a consequence it has been [1:05] possible to achieve the transition into [1:08] a convoy driving [1:10] mode since the early 1990s it has also [1:13] been possible to perform Lane changes [1:17] autonomously in 1993 94 in the framework [1:21] of the third phase of the Eureka project [1:24] Prometheus three Mercedes-Benz S-class [1:27] Passenger cars were equipped with [1:29] University bundas ver Vision systems [1:32] based on new micr miniaturized TV [1:36] cameras all electronic equipment now [1:39] fits into the trunk space in order to [1:41] achieve better maintainability and [1:44] accessibility however in the ubm vehicle [1:47] Vamp part of it has been mounted on the [1:50] right back [1:52] seat up to six other vehicles may be [1:55] tracked in parallel in both the front [1:58] and the rear hemisphere [2:00] their state relative to the ego vehicle [2:03] is estimated concurrently at a rate of [2:05] 25 [2:07] Hertz on this basis at the final [2:09] demonstration of the Prometheus project [2:11] near Paris in October of 1994 on a [2:14] three-lane freeway with normal traffic [2:17] the capability of making the decision [2:19] for lane change fully autonomously was [2:22] shown for the first [2:24] time a module developed by our computer [2:27] science Partners at ubm tracks the [2:29] entire traffic situation around the [2:31] vehicle especially in the neighboring [2:33] Lane to be changed to