Motional 与 MIT 合作研究让自动驾驶汽车实时解释决策

Motional：https://motional.com/ 和 MIT：https://www.mit.edu/ 的研究人员开发了一种系统，使自动驾驶汽车能够实时解释其决策，从而解决自动驾驶车辆 AI 的“黑箱”问题。

这项工作发表在 Nature：https://www.nature.com/ 上，由 Motional 团队完成，团队成员包括 CEO Laura Major，并与 MIT 计算机科学与人工智能实验室的研究人员合作。他们提出的方法被称为概念包装网络（Concept-Wrapper Network，简称 CW-Net），旨在将自动驾驶系统神经网络的内部计算转化为人类可以理解的概念。

如果一辆现有的自动驾驶汽车在没有明显危险的空旷道路上紧急刹车，驾驶员和乘客都无法知道原因。现代自动驾驶系统越来越依赖于基于大量驾驶数据训练的神经网络。这些网络性能可能很好，但它们不会公开其推理过程，因此工程师将其形容为黑箱。

CW-Net 的工作原理是将自动驾驶系统的内部逻辑转化为诸如“接近静止车辆”或“靠近骑行者”等概念。据 Motional 介绍，这些概念可以显示在仪表盘上，实时展示哪些概念在影响车辆的驾驶决策。

该系统的设计使解释不是事后生成的对网络行为的推测，而是车辆的最终决策系统直接基于这些可被人类理解的概念采取行动，因此一次刹车事件可以追溯到触发该动作的具体概念。Motional 将此描述为因果上可信的，与那些生成自然语言解释的方法不同，后者可能看起来合理，但不一定准确。

Laura Major 在强调这种可解释性时，将其与纯粹依赖端到端深度学习处理驾驶决策的替代方案进行了对比。

“纯端到端的方法通常可以达到非常好的 80%–90% 的——甚至可能达到 95%——解决方案，但这还不足以让驾驶员下车或赢得城市、社区和客户的信任，”她说。

根据Motional的说法，可解释的人工智能研究在很大程度上一直局限于实验室环境中的计算机模拟。Motional和MIT团队则将CW-Net部署在一辆有经验的安全驾驶员坐在驾驶席的自动驾驶车辆上，在私人测试赛道以及拉斯维加斯周围的公共道路上收集数据。

团队使用了其基于深度学习的规划系统的早期实验版本，据称该系统显示出了有竞争力的性能，但存在明显的不足，而CW-Net可以帮助揭示这些问题。测试中的两个事件说明了系统捕捉到了什么。

在其中一次测试中，自动驾驶车辆在交通锥附近反复停车，车辆操作员认为是交通锥本身触发了这种行为。研究人员移走交通锥，但车辆仍然停了下来。CW-Net的显示显示了实际原因：该实验性规划系统在前方“产生幻觉”，以为有一辆车停在那里，这种模式可以追溯到其训练数据。这个解释让研究人员理解、预测并解决了问题。

第二次测试涉及一名骑自行车的人。自动驾驶车辆如预期般检测到并为骑车者停车，但CW-Net显示实验性规划系统实际上并没有基于骑车者的存在作出决策。安全驾驶员在注意到这一点后，在遇到骑车者时采取了更加谨慎的应对。后续分析证实保持谨慎是必要的，因为当时车辆的刹车是由安全备用系统而非实验性的深度学习规划器触发的。

为AI系统增加可解释性的层次会带来速度和性能上的已知代价，Motional也承认这一风险。然而，当研究人员将CW-Net与领先的自动驾驶算法进行基准测试时，驾驶能力的差异不到1%。

拉斯维加斯的事件显示了这种权衡在操作上而不仅仅是在学术上为何重要。一名能够看到停车是由“幻觉车辆”引起，或者由备用系统而非主规划器负责的安全驾驶员，可以比仅依靠行为观察的驾驶员更精确地做出反应和报告。

这种可见性直接影响工程团队诊断系统的速度，以及安全操作员区分预期行为和故障的信心。

Motional 将 CW-Net 的工作与自动驾驶汽车运营商在技术扩展到新市场和辖区时所面临的更广泛压力联系起来。监管机构自然会要求对 AI 系统如何做出决策提供更多透明度，并预计像 CW-Net 这样的工具可能会从研究项目转向成为基本要求。

除了乘用车，自动无人机甚至机器人手术也被列为其他安全关键领域，在这些领域中，操作员和开发人员将需要理解系统的能力、局限性和意外行为的方法。

在 Physical AI Expo 了解更多关于物理 AI 的信息：https://physicalaiconference.com/，该展会在阿姆斯特丹、伦敦和北美举办。

另请参见：MIT AI 在没有历史数据的情况下预测极端天气：https://www.artificialintelligence-news.com/news/mit-ai-forecasts-extreme-weather-without-historical-data/

想要向行业领先者了解更多关于 AI 和大数据的信息吗？请查看 AI & Big Data Expo：https://www.ai-expo.net/?utm_source=AI-News&utm_medium=Footer-banner&utm_campaign=world-series，该展会在阿姆斯特丹、加利福尼亚和伦敦举办。该综合性活动是 TechEx 的一部分：https://techexevent.com/?utm_source=AI-News&utm_medium=Footer-banner&utm_campaign=world-series，并与其他领先的技术活动共同举办，包括 Cyber Security & Cloud Expo：https://cybersecuritycloudexpo.com/?utm_source=CloudTech-News&utm_medium=Footer-banner&utm_campaign=world-series。点击这里了解更多信息：https://techexevent.com/?utm_source=AI-News&utm_medium=Footer-banner&utm_campaign=world-series

AI 新闻由 TechForge Media 提供动力：https://techforge.pub/?utm_source=AI-News&utm_medium=Footer-banner&utm_campaign=world-series。 在此探索其他即将举行的企业技术活动和网络研讨会：https://techforge.pub/events/?utm_source=AI-News&utm_medium=Footer-banner&utm_campaign=world-series。

Ryan Daws：https://www.artificialintelligence-news.com/news/author/ryan/

零售与物流人工智能：https://www.artificialintelligence-news.com/categories/ai-in-action/retail-logistics-ai/，服务行业人工智能：https://www.artificialintelligence-news.com/categories/ai-in-action/service-industry-ai/

物理人工智能：https://www.artificialintelligence-news.com/categories/ai-and-us/physical-ai/

人工智能商业策略：https://www.artificialintelligence-news.com/categories/inside-ai/ai-business-strategy/, 人工智能：https://www.artificialintelligence-news.com/categories/artificial-intelligence/, 特写：https://www.artificialintelligence-news.com/categories/features/, 金融人工智能：https://www.artificialintelligence-news.com/categories/ai-in-action/finance-ai/, 职场世界：https://www.artificialintelligence-news.com/categories/ai-and-us/world-of-work/

AI 实践：https://www.artificialintelligence-news.com/categories/ai-in-action/

将我们所有的高级内容和最新科技新闻直接发送到您的邮箱

AI 新闻是 TechForge 的一部分：https://techforge.pub/

Motional：https://motional.com/ and MIT：https://www.mit.edu/ researchers have built a system that lets self-driving cars explain their decisions in real-time, tackling the black-box problem in autonomous vehicle AI.

The work, published in Nature：https://www.nature.com/, comes from a team at Motional that includes CEO Laura Major, working alongside researchers from MIT’s Computer Science and Artificial Intelligence Laboratory. Their proposed method, called the Concept-Wrapper Network or CW-Net, aims to translate the internal calculations of a self-driving system’s neural network into concepts a human can actually read.

If a current self-driving car brakes hard on a clear road with no obvious hazard in sight, neither the driver nor a passenger has any way of knowing why. Modern self-driving systems increasingly rely on neural networks trained on large volumes of driving data. Those networks can perform well, but they don’t expose their reasoning, which is why engineers describe them as black boxes.

CW-Net works by converting a self-driving system’s internal logic into concepts such as “Approaching Stopped Vehicle” or “Close to Cyclist.” These could, according to Motional, appear on a dashboard showing which concepts are influencing the vehicle’s driving decisions as they happen.

The system is designed so the explanations aren’t generated after the fact as a guess at what the network might have been doing. Instead, the vehicle’s final decision-making system takes action based directly on these human-interpretable concepts, so a braking event traces back to a specific concept that triggered it. Motional describes this as causally faithful, distinguishing it from approaches that generate natural-language explanations, which can read as plausible without necessarily being accurate.

Laura Major frames the case for this kind of interpretability against the alternative of relying purely on end-to-end deep learning to handle driving decisions.

“The general end-to-end only approach can get to a really good 80-90 percent – maybe even 95 percent – solution, but that’s not good enough to remove a driver or to earn the trust of cities, communities, and customers,” she said.

Explainable AI research has largely stayed confined to computer simulations in lab settings, according to Motional. The Motional and MIT team instead deployed CW-Net on an autonomous vehicle with an experienced safety operator in the driver’s seat, collecting data on a private test track and on public roads around Las Vegas.

The team used an earlier experimental version of its deep-learning-based planning system, described as showing competitive performance but with notable shortcomings that CW-Net could help surface. Two incidents from the testing illustrate what the system caught.

In one, the autonomous vehicle repeatedly stopped near a traffic cone, and the vehicle operator assumed the cone itself was triggering the behaviour. Researchers removed the cone and the car stopped anyway. CW-Net’s display showed the actual cause: the experimental planning system was hallucinating a stopped vehicle ahead, a pattern traced back to its training data. That explanation let the researchers understand, predict, and resolve the issue.

A second test involved a cyclist. The autonomous vehicle detected and stopped for the cyclist as expected, but CW-Net revealed that the experimental planning system wasn’t actually basing its decision on the cyclist’s presence. The safety driver responded by exercising more caution around cyclists after noticing this. Follow-up analysis confirmed that caution was warranted, because the vehicle’s braking in that case came from a safety backup system rather than the experimental deep-learning-based planner.

Adding layers of explainability to an AI system carries a known cost in speed and performance, and Motional acknowledges that risk. However, when researchers benchmarked CW-Net against leading autonomous driving algorithms, the difference in driving capability came in at less than one percent.

The Las Vegas incidents show why that trade-off matters operationally rather than just academically. A safety driver who can see that a stop is caused by a hallucinated vehicle, or that a backup system rather than the primary planner is responsible for a manoeuvre, can respond and report with more precision than one working from behaviour alone.

That visibility feeds directly into how quickly an engineering team can diagnose a system, and how confidently a safety operator can distinguish between an intended behaviour and a fault.

Motional connects the CW-Net work to broader pressure on autonomous vehicle operators as the technology extends into new markets and jurisdictions. Regulators are naturally asking for more transparency about how AI systems reach their decisions, and it expects tools like CW-Net could move from research projects toward a baseline requirement.

Beyond passenger vehicles, autonomous drones and even robotic surgery are cited as other safety-critical domains where operators and developers will need ways to understand a system’s capabilities, limitations, and unexpected behaviours.

Learn more about physical AI during the Physical AI Expo ：https://physicalaiconference.com/ held in Amsterdam, London, and North America.

See also: MIT AI forecasts extreme weather without historical data ：https://www.artificialintelligence-news.com/news/mit-ai-forecasts-extreme-weather-without-historical-data/

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo：https://www.ai-expo.net/?utm_source=AI-News&utm_medium=Footer-banner&utm_campaign=world-series taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx：https://techexevent.com/?utm_source=AI-News&utm_medium=Footer-banner&utm_campaign=world-series and is co-located with other leading technology events including the Cyber Security & Cloud Expo：https://cybersecuritycloudexpo.com/?utm_source=CloudTech-News&utm_medium=Footer-banner&utm_campaign=world-series. Click here：https://techexevent.com/?utm_source=AI-News&utm_medium=Footer-banner&utm_campaign=world-series for more information.

AI News is powered by TechForge Media：https://techforge.pub/?utm_source=AI-News&utm_medium=Footer-banner&utm_campaign=world-series. Explore other upcoming enterprise technology events and webinars here：https://techforge.pub/events/?utm_source=AI-News&utm_medium=Footer-banner&utm_campaign=world-series.

Ryan Daws：https://www.artificialintelligence-news.com/news/author/ryan/

Retail & Logistics AI：https://www.artificialintelligence-news.com/categories/ai-in-action/retail-logistics-ai/, Service Industry AI：https://www.artificialintelligence-news.com/categories/ai-in-action/service-industry-ai/

Physical AI：https://www.artificialintelligence-news.com/categories/ai-and-us/physical-ai/

AI Business Strategy：https://www.artificialintelligence-news.com/categories/inside-ai/ai-business-strategy/, Artificial Intelligence：https://www.artificialintelligence-news.com/categories/artificial-intelligence/, Features：https://www.artificialintelligence-news.com/categories/features/, Finance AI：https://www.artificialintelligence-news.com/categories/ai-in-action/finance-ai/, World of Work：https://www.artificialintelligence-news.com/categories/ai-and-us/world-of-work/

AI in Action：https://www.artificialintelligence-news.com/categories/ai-in-action/

All our premium content and latest tech news delivered straight to your inbox

AI News is part of TechForge：https://techforge.pub/