Self-cultivation of an intelligent driving product manager

So, what is the difference between product managers for intelligent driving and product managers in other fields? What are the goals and specific content of its work? What are the requirements for practitioners? How can I be a good product manager for smart driving?

With these questions, this article will share with you some of the author’s insights and experiences in working as a smart driving product manager for many years.

Before understanding the work content of an intelligent driving product manager, we need to first know the current routine process of intelligent driving function development. It should be noted that the current smart driving product managers are mainly responsible for the functions and effects of smart driving. Therefore, the smart driving products mentioned in this article are approximately equal to smart driving functions.

△Figure 1 V-shaped process of intelligent driving function development

△Figure 2 Intelligent driving development team

The current development of intelligent driving generally follows a V-shaped development process, as shown in Figure 1. Starting from functional requirements analysis, it is gradually decomposed into software module codes, and then gradually integrated and tested, and finally realizes various functions of intelligent driving. It should be noted that each step of the process in the figure is not necessarily in strict sequence. For example, component analysis and design and software analysis and design are often carried out simultaneously.

Corresponding to the development process in Figure 1, the development team and division of labor in Figure 2 are formed. Among them, Zhijia’s product manager and the product team are responsible for the first step in the development process: functional requirements analysis.

It can be seen that the main work goal of an intelligent driving product manager is to analyze the functional requirements of intelligent driving and pass the clear functional requirements to the downstream as the development goal. The functional requirements mentioned here specifically include two aspects: first, what functions the smart driving product needs to implement and what application scenarios it needs to meet; second, what is the specific performance of each function and the effect it can achieve.

It can be said that the intelligent driving product manager is the navigator of intelligent driving function development, responsible for clarifying the development goals and guiding the development direction.

So, in order to achieve work goals, what specific work content does a smart driving product manager need to do? Let’s take a look at the common requirements for smart driving product managers on recruitment websites - Figure 3 shows three job descriptions for smart driving product managers that we found on a well-known recruitment website.

It is not difficult to find that the work content of smart driving product managers is mainly product planning and product definition, which is consistent with the two aspects of functional requirements mentioned earlier: product planning, planning which functions are to be implemented and which scenarios are to be solved. ; Product definition defines the specific performance of functions, as well as performance and effects. In addition, smart driving product managers need to constantly study industry and market trends in order to formulate appropriate product roadmaps and make reasonable product planning and definitions.

△ Figure 3 Recruitment requirements for smart driving product managers

Let’s explain in detail what product planning, product definition, market research, etc. include.

(1) The planning of smart driving products mainly involves planning: which functions need to be developed, which scenarios should be used, and what user pain points should be solved. Product planning outputs a roadmap for smart driving products, that is, a roadmap, which is the short, medium, and long-term planning for the company's smart driving products. For example, at a specific point in time, what smart driving functions need to be equipped, and what can What level of intelligent driving has been achieved (L2/L3/L4).

The picture below is a history of the development of Tesla’s smart driving products, which can be regarded as a simplified version of Roadmap. Tesla's intelligent driving includes 4 generations of platforms, each generation of platform is equipped with new functions; from early simple functions such as forward collision warning, through continuous upgrades and iterations, it finally achieves fully autonomous driving.

△Figure 4 Tesla’s intelligent driving development history

(2) The definition of smart driving products outputs a product requirement document, that is, Product Requirement Document (PRD), which includes a description of a certain function, the basis for product design, hardware configuration requirements, the scope of application of the function, and application scenarios. Function implementation process, detailed functional requirements definition, human-computer interaction requirements, experience evaluation indicators, data requirements, iteration plans, etc. PRD is the main work output content of product managers. It is the source and guiding document for smart driving product development. It is an important basis for downstream departments to develop system architecture and algorithms.

Generally speaking, a certain intelligent driving function can form a PRD, such as the PRD of the adaptive cruise ACC function, the PRD of the high-speed pilot assist NOA function, etc. But sometimes, in order to reflect the completeness and consistency of smart driving products, all smart driving functions will be summarized into a PRD, which will subdivide each function and the relationship between functions.

(3) Market research is to understand the situation of products and markets, so as to make more reasonable, competitive and innovative product planning and product definitions. Market research is an in-depth investigation and study of the market status, development, trends, etc. of intelligent driving, including similar products, related technologies, policies, capital, users, etc. Typical outputs of smart driving market research include competitive product benchmarking reports, user survey questionnaires, interpretations of emerging technologies, policy analysis, investment and financing status of similar companies, etc.

In the early stage of Jiuzhang Smart Driving, there were some evaluation articles on mainstream smart driving products, which are typical competitive product analysis and research, such as "Evaluation of Automatic Parking Products of Tesla, Xpeng, and NIO" and "Tesla, NIO, "Evaluation of Xpeng's Three Mainstream Intelligent Driving Products"; in addition, many professional media such as No. 42 Garage, Cheyun, etc. also have evaluations and analyzes of intelligent driving products, and their content can also be used as a reference for competitive product analysis.

Figure 5 shows some user survey results about intelligent driving. In addition, tracking and research on technology trends, regulations, policies and capital dynamics are also important aspects of market research.

△Figure 5 User survey of intelligent driving

(Data sources: Audi, American Automobile Association, Bitauto Research Institute)

As the saying goes, "There is a beginning and an end." As a product manager, it is naturally the same. As the owner of the product, the product manager must accept the final status and effect of the product, confirm the performance and effect of the product, and achieve the status required by the PRD. The professional term for this acceptance process is "walk-through".

In an ideal state, walkthrough is the last step of the entire development process, which is to confirm the results of product development and testing; however, in the actual development process, due to objective factors such as technical level and multi-department communication costs, downstream departments It is difficult to develop in accordance with PRD 100%, so product managers usually need to intervene in advance for inspections and start gradual acceptance during the software testing stage to ensure that various functions can achieve the expected results.

The above is the work content of the intelligent driving product manager, including all aspects related to the product such as planning, management, definition, design, and iteration of intelligent driving products, as well as the market, industry, users, policies, capital, etc. Research and analysis. To do these tasks well, there must be high requirements for the professional skills and knowledge reserves of smart driving product managers. So, how to do a good job as a smart driving product manager? What knowledge and skills are required? What kind of person is suitable and capable? We also summarized and interpreted it.

根据前面介绍的智驾产品经理工作内容，我们可以看出，这是一个对知识储备和经验方法都有很高要求的岗位。一名优秀的智驾产品经理，需要全方位地掌握智能驾驶相关的知识，并且要具备一套适合智能驾驶开发的产品经理方法论。下面我们具体展开来说。

智能驾驶作为新兴的领域和方向，是人工智能、互联网、大数据与传统汽车工程等多学科的交叉和深度融合，因此对从业人员的知识储备要求较高，尤其是作为领航员的产品经理，更是要熟悉智能驾驶涉及到的方方面面。

从字面意思分开来看：智能驾驶=智能+驾驶。所以，智驾的产品经理，需要对智能和驾驶（汽车工程）这两个维度的知识，都有深刻的理解。

（1）智能化相关知识。 提到智能化，首先想到的就是电子化、信息化、人工智能、算法、机器学习、芯片等。的确，智能驾驶也与这些紧密相关。对于智能驾驶的三大模块：感知定位、决策规划、控制执行来说，每个模块都离不开智能化。

△图6 智能驾驶的三大模块

感知定位模块通过各种传感器来识别周围环境，并通过地图、导航确定车辆的位置。其中，基于摄像头的目标检测与识别，需要神经网络和机器学习算法的加持；激光雷达的点云数据处理，涉及到聚类算法，这些都属于人工智能的范畴。另外，机器学习需要基于大量的数据，不断优化提升，这又涉及到大数据的分析和处理。

△ 图7 摄像头的目标检测与识别

（图片来源： http://www.cheyun.com/content/40243 ）

△图8 激光雷达的点云数据

（图片来源： https://www.sohu.com/a/208301629_11570 ）

决策规划模块通过感知定位模块传输的信息，分析计算，做出决策，然后规划车辆应该行驶的轨迹。其中的任务决策和轨迹规划两大块内容，自然也是要基于各类算法来实现。

控制执行模块虽然只是接收指令，执行任务，但横向控制和纵向控制并不是孤立的，他们之间也存在一定的耦合关系。如何协调车辆的横向与纵向运动，做到安全舒适，也离不开各项动力学参数的智能化匹配。

此外，智能驾驶往往与网络通信密不可分，尤其是对于车路协同方案来说，V2X更是离不开网络通信技术。车-车、车-人、车-设备、车-云，信息间的连通，都属于互联网、物联网，或者车联网的范畴。

△图9 V2X的细分内容

（图片来源： https://www.sohu.com/a/389494149_120658453 ）

（2）汽车工程知识。 智能驾驶可以看作是搭载在汽车上的人工智能，因此“驾驶”的部分，也就是汽车本身，也至关重要。自然，智驾产品经理，也要懂得汽车理论与构造相关的知识内容。

以量产的乘用车来说，主要由动力系统、底盘系统、车身、内饰、外饰和电子电器系统组成。其中，随着近几年汽车智能化的发展，内饰中的仪表、车机等电器设备逐渐占据了重要地位，与其他内饰部件一起，组成了座舱系统。

△ 图10 乘用车的系统构成

（图片来源： https://www.sohu.com/a/326111904_157534 ）

智能驾驶涉及到的主要是电子电器、动力、底盘、座舱4个部分，其中电子电器系统作为智能驾驶的直接载体，需要重点关注：软件方面，电子电器架构直接决定了汽车的智能化水平，从传统的分布式，到特斯拉的域集中式，再到目前最先进的中央集中式架构；硬件方面，智能驾驶涉及到的各类传感器、芯片、控制器，也需要产品经理有所了解，包括其工作原理、主要参数等等。

△图11 车辆电子电器架构的演变

（图片来源： https://zhuanlan.zhihu.com/p/416435460 ）

△表1 某款域控制器的参数

座舱是用户与汽车直接交流的系统，座舱中的仪表、车机、灯光、方向盘、座椅、安全带等，都可以与用户形成有效的人机交互。人机交互作为智能驾驶的一个重要环节，直接影响用户对智能驾驶功能的使用体验，因此十分重要。然而，人机交互，Human–Machine Interaction，也是一门非常复杂和的工业设计专业学科，有一系列设计理论和方法，也是需要产品经理熟悉的内容。

再来说说汽车的理论。汽车专业背景的同行都知道，一款车的性能，通常可以从动力学、经济性、制动性能、稳定性、平顺性、通过性等6个方面来综合评价。对于智能驾驶来说，其中的稳定性和平顺性尤为重要，这两项性能，体现出智能驾驶带给用户的舒适程度。作为智驾的产品经理，需要熟知影响稳定性和平顺性的主要参数是什么，比如纵向加减速度、横摆角速度、侧向加速度等等，以及这些参数如何影响车辆的性能。

作为智能驾驶的产品经理，需要有一套完善的、有效的方法论，才能设计和打磨出优秀的智能驾驶产品。这一套方法论，涵盖了做智驾产品的思维方式、流程规则和具体的方案等。

（1）互联网思维和工程化思维的融合。

产品经理是源自互联网行业的岗位，智能驾驶又是人工智能与汽车工程深度结合的产物，因此作为智能驾驶产品经理，既要有互联网的思维模式，又要有传统汽车行业的工程化思维模式，并且能在实际的项目中，将两种思维模式融合起来，才能做出能够工程化落地的有竞争力的智驾产品。

互联网产品突出的是迭代、极致与跨界。迭代，就是允许不足、不断试错，在持续的产品迭代升级中，逐渐完善产品，耳熟能详的“敏捷开发”就是基于迭代的一种典型的互联网开发流程。极致，就是把产品和用户体验做到最好，不放过每一个细节，提供超出用户预期的产品，解决用户的所有痛点。跨界，就是打破固有的边界，万物互联，建立生态体系。

传统汽车行业重视的是工程化落地。作为高度精密的工业产品，最重要的是能够工程化地生产，能够保证高度安全可靠，因此不允许试错，不允许为了某个细节而影响整体的平衡。传统的汽车产品是稳定的，不会有快速的迭代，也不会有过多与外界的联通和交互。任何的创新与设计，都需要充分考虑到工程化和批量化的可行性。

智能驾驶的产品兼具互联网产品和传统汽车产品的双重属性。一方面，智能驾驶涉及到先进的算法、海量的数据、无穷无尽的场景，并且用户的体验直接影响产品的效果，因此需要采用互联网的思维模式来做；另一方面，智能驾驶始终是汽车的一部分，需要保证安全性、可靠性，并且容易工程化落地，实现批量生产，因此需要采用工程化的思维模式来做。

看起来截然不同的两种思维方式，我们该如何去融合呢？这是一个很复杂的问题，并且对不同的案例，会有不同的方案，不能一概而论。总体来说，这是一个权衡利弊、动态博弈的过程，是敏捷开发流程和传统瀑布式开发流程互相取长补短的过程，也考验着产品经理对智能驾驶产品开发的理解。我们举几个例子：

对于直接搭载在车上，与整车的设计制造强相关的部分，例如各类传感器和芯片等硬件，需要严格按工程化的方式来设计和开发。在设计初期就考虑到全生命周期的各种情况，形成可靠的平台化方案。

电子电器架构和底层软件部分，虽然不是直接装配在车上的硬件，但是智能驾驶软件开发的基础和框架，应该在初期就全面考虑，尽量减少后期的变更；但为了保持一定的灵活性，我们可以在架构和底层软件的某些小模块中，预留一些接口，保留更多的可能性。

应用层的算法，是建立在底层软件、中间件的基础之上的，是一个个模块，其单独的变更，不影响整体的系统架构，也不会涉及其他模块的变更，因此在满足安全要求的前提下，可以快速迭代、不断优化，通过OTA的方式逐步升级，允许一定的试错成本。

（2）对出行场景的分析。

智能驾驶解决的是用户对于出行场景的智能化需求，应用的场景就是各种各样的车辆出行场景，因此作为智驾产品经理，对出行场景的分析十分重要。可以说，分析清楚了出行场景，就完成了一半的智能驾驶产品设计。

一套系统化、完整化的出行场景库，是智能驾驶产品设计与定义的基础。对于出行场景的分析，其实每家公司都有自己的一套方法，比如基于基本交通元素的泛化推演、基于用户日常出行的数据统计、基于经验和想象得出的常规场景等等。不同方法得出的出行场景库，是各不相同的，很难说孰优孰劣。这里简单介绍一套我们自己的场景分析方法，为读者提供一定的思路。

首先，出行场景主要分为行车场景和泊车场景。行车场景就是在公共道路上行驶时，所遇到的各类场景；泊车场景就是在停车场内遇到的各类场景。

行车场景按照道路的开放与复杂程度，又分为高速公路、城市快速路、城区道路、城际道路、乡村道路等，这些不同类型的公共道路，都有其各自的特点，需要充分考虑其特有的场景。

比如，高速公路和城市快速路是高度结构化的道路，有很强的规则特征；并且相对封闭，除了出口和入口外，基本没有别的通道，路上没有行人、两轮车这些移动的障碍物，减小了不确定性，可以认为是一段理想的封闭环境。对于这两种道路来说，匝道场景是其特有的场景，需要重点研究。

再比如，城区道路属于开放的交通环境，交通元素复杂，不仅有各种车道线、交通标志，还有机动车、非机动车等交通参与者，不仅有规则的主干道路，也有非常不规则的各种辅路、窄路等，有着高度的不确定性。城区道路的典型场景是十字路口，环岛，公交站台，学校、小区、商场门口等各类场景。

此外，还有一些场景，是在任何类型的公共道路上，都存在的，例如车道内行驶、变道等。我们可以作为一类场景考虑。

泊车场景主要是各类的停车场和各种各样的停车位。常见的停车场类型有地下停车库、露天停车场、停车楼等；停车位则涉及各类车位线的样式、车位附近的参照物、障碍物等等。对于泊车场景来说，主要是各种基本元素的排列组合，形成各种各样的场景。

△表2 出行场景

（3）对用户体验的分析

任何产品，面对的对象都是终端用户，产品的好坏也是由用户来评价，因此用户体验是产品设计和开发的第一目标。可见，能够精准地分析用户体验，从用户的角度来设计智能驾驶的产品方案，是非常重要的。

目前对于智能驾驶的评价，往往基于工程化的性能参数和工程师想象出来的各种场景，编制出测试用例，然后基于测试用例来测评智能驾驶的产品表现。在这一过程中，所有的测试评价内容，都是从工程师或者开发者的角度来考虑问题，往往会忽略一点：产品最终的使用对象是普通用户。普通用户的专业知识和思考方式，和智能驾驶从业者是不一样的，如果没有意识到这一点，那么所开发出来的智能驾驶很容易变成开发者自嗨的方案，用户很可能不买单。

一些业内人士司空见惯的方案，可能用户并不喜欢，这就是需要产品经理去优化和改进的。想要做好用户体验，产品经理应该把自己真正当成用户，以非智能驾驶从业者的身份，去琢磨智驾产品应该如何设计，从用户使用智能驾驶的维度，来思考和定义，跳出既往的思维定势和各种技术限制，用优化和发展的眼光去设计智能驾驶的方案。

下面分享2个提升用户对智能驾驶的体验的经典案例。

a. 主动安全功能开关的优化：

几乎每家车企，都把传统的主动安全功能，作为智能驾驶的一部分。主动安全功能多种多样，比如前向碰撞预警FCW、自动紧急制动AEB、盲区检测BSD、变道预警LCW等等。我们统计了市面上已有的主动安全功能，大约有15～20种，而每项功能往往都需要一个开关来控制。

这样就导致一个问题：如果车上搭载了全方位的主动安全功能，那么就会有非常多的开关，占据大量的HMI显示空间，并且使用频率低，便捷性差；同时增加了用户的学习成本，作为普通用户，没有必要去理解这些专业词汇所代表的含义，去理解每个功能的详细内容，他们所希望的，就是能简单地使用而已。

针对这一问题，我们调研了一些车主的需求后，做出了一个大胆的创新设计：主动安全功能的开关设置，不再以单独的功能来区分，而是根据所防护的区域和方位，来做开关设置。这样，优化后的开关设置就只有“前方安全守护”“后方安全守护”“侧前方安全守护”“侧后方安全守护”等，在简化开关的同时，也省去了用户对功能描述的学习成本，更加直接简单地选择所需要的功能，也更符合普通用户的使用习惯。

△图11 主动安全的开关设置

b. 自适应巡航车速设置的优化：

自适应巡航ACC（Adaptive Cruise Control），能够控制车辆，按用户设定的车速行驶；用户可以随时设置车速，系统会按新的车速来控制车辆的加减速。

The ACC function is a very mature function, and many models are already equipped with it. We found that in any current car model, the user's speed settings are basically two-speed adjustment: 10 or 5kph as a step, and 1kph as a step. From the perspective of engineering development, it is necessary to retain 1kph as the step size to achieve the maximum vehicle speed setting accuracy. However, in the actual user use process, statistical results show that almost no users have used 1kph to adjust the speed. Basically, they adjust the speed by 10 or 5kph. This is also consistent with common sense: when driving on the road, users do not care about the speed of 1kph. Changes, usually only changes in vehicle speed of 5 or 10kph, are meaningful.

Based on the above reasons, we broke the consistent practice in the industry and set up two speed setting methods with steps of 10kph and 5kph respectively, and canceled the method of 1kph as step; although the accuracy is lower, the actual use experience and Efficiency will be improved and users will be more "enjoyed".

There was a very popular saying in the past few years: "Everyone is a product manager." But for intelligent driving, not everyone can be a product manager. As a multi-disciplinary field, intelligent driving must achieve a balance in multiple dimensions such as safety, experience, and technology. Compared with flexible and changeable Internet products, it must meet safety and technical requirements, especially the requirements of vehicle regulations; Compared with traditional vehicle development, we must pay more attention to user experience and innovation, and cannot be limited to the engineering mindset. Therefore, smart driving product managers need to possess some personal characteristics in order to design and lead satisfactory smart driving products.

(1) Innovation ability. The time for the development of intelligent driving is limited, especially for mass-produced intelligent driving products, which are a new species that was only born in the past ten years. Therefore, there are still strong technical limitations and many shortcomings. Product managers need to be good at thinking, boldly innovate, jump out of existing technical limitations and thinking stereotypes, and constantly improve and optimize. It can be said that innovation is the core and essential quality of smart driving product managers.

(2) Safety awareness. Since smart driving is implemented in driving, safety must be the most important. All smart driving products and functions must put safety first. For smart driving, safety is a veto item and a red line. No matter how good the user experience is and how advanced the technology is, as long as the safety is not enough, it needs to be modified and improved. As a product manager for intelligent driving, you must have a strong sense of safety, be familiar with traffic regulations, and be familiar with various dangerous scenarios and solutions to ensure user safety to the greatest extent. In addition, car-level standards are also a part of safety. Only solutions that meet the requirements of car regulations are safe solutions; therefore, smart driving product managers must also have a basic concept of car regulations.

(3) Awareness of responsibility. The product manager is the owner of the product and is responsible for the final status and effects of the product. Although the development process of intelligent driving involves many departments, product managers still need to check every link, from system architecture, to software and hardware module development, to software testing, integration testing, real vehicle testing, etc., to confirm that the product is running in accordance with the requirements Your own design plan is implemented. You must always regard yourself as the first person responsible for the product and the owner, and follow up the entire process.

( 4) Keep your feet on the ground. As mentioned earlier, product managers must have sufficient innovation capabilities and be able to look up to the stars. However, innovation does not mean flying in the sky, but rather proposing achievable solutions and being down-to-earth. For example, the ultimate form of autonomous driving is L5 fully autonomous driving. However, due to current technology, policy and other restrictions, it is impossible to reach the L5 level in the short term; so in the short term, we can propose many human-machine Shared driving solutions, however, cannot require the realization of fully autonomous driving in all scenarios in the short term. If someone says that L5 level autonomous driving will be achieved within 3-5 years, it must be unreliable.

The above is some of our experience summary and insights about intelligent driving product managers. From our sharing, it is not difficult to find that the product manager of Zhijia is a position that requires high comprehensive quality and knowledge accumulation. It requires not only the personal characteristics of looking up to the stars and being down-to-earth, but also having sufficient deep knowledge and experience. method. But precisely because of this, excellent smart driving product managers are particularly rare. We hope that product managers in the industry can communicate more, share more, brainstorm, and jointly promote the progress of intelligent driving products and improve user satisfaction.