SBIR Phase II: Human agent prediction for autonomous vehicles

小企业创新研究(SBIR)第二阶段项目的更广泛影响/商业潜力源于这样一个事实，即它将释放密集城市环境中自动驾驶车辆的潜力，使这些车辆将安全、有效，并能够在有广泛易受伤害的道路使用者的环境中运行。通过为自动驾驶车辆提供一个系统来理解人类在道路上的目标和行为，这项技术将使自动驾驶车辆能够安全有效地对人类做出反应。如果没有在该奖项的帮助下将创新商业化，自动驾驶汽车充其量将是毫无用处的胆小和危险的城市道路上的补充，充其量是实现安全、宜居城市目标的致命障碍。每一辆有能力投放市场的自动驾驶汽车都需要解决这个项目正在帮助解决的问题。这个小企业创新研究(SBIR)第二阶段项目将有助于解决自动驾驶汽车中最棘手的问题之一。计算机系统如何才能了解人类的心理状态，这一问题自这个术语诞生以来，就一直困扰着对机器智能感兴趣的研究人员和外行。通过建立一种基于利用仔细的人类测量和最先进的学习算法的方法，该项目中开发的创新将有助于铺平一条新的道路，用于对认知设施进行计算建模，这些设施是人类智能的核心，但历来难以使用传统的机器学习或计算机视觉技术进行建模。除了帮助解决人类对自动驾驶汽车的理解这一核心问题外，该项目发表的研究还将为理解一大类问题开辟新的途径，在这些问题中，关于世界的“地面真相”问题很难或不可能得到回答。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase II project results from the fact that it will unlock the potential of autonomous vehicles in dense urban environments in such a way that these vehicles will be safe, effective, and able to operate in environments with a wide range of often-vulnerable road users. By providing a system for autonomous vehicles to understand the goals and behaviors of humans on the road, the technology will allow autonomous vehicles to react to humans safely and effectively. Without the innovations commercialized with the help of this award, autonomous vehicles will be at best uselessly timid and dangerous additions to urban roads, and at worst a deadly obstacle to the goal of safe, livable cities. Every autonomous vehicle that is capable enough to be on the market will need to solve the problem that this project is helping to solve. This Small Business Innovation Research (SBIR) Phase II project will help to address one of the thorniest problems in autonomous vehicles. The question of how a computer system can gain an understanding of human mental states has occupied researchers and laypeople with an interest in machine intelligence since the coining of the term. By building an approach based on the leveraging of careful human measurement and state-of-the-art learning algorithms, the innovations developed in this project will help pave a new path towards the computational modeling of cognitive facilities that are central to human intelligence but historically intractable to model using conventional machine learning or computer vision techniques. In addition to helping solve the central problem of human understanding for autonomous vehicles, the research published from this project will open new avenues for understanding a broad class of problems where the question of "ground truth" about the world is difficult or impossible to answer.