Symbolic dynamics and model-based control of autonomous vehicles

自动驾驶车辆的符号动力学和基于模型的控制

• 批准号: 506419-2016
• 负责人: McPhee, John
• 金额: $ 3.64万
• 依托单位: University of Waterloo
• 依托单位国家: 加拿大
• 项目类别: Collaborative Research and Development Grants
• 财政年份: 2018
• 资助国家: 加拿大
• 起止时间: 2018-01-01 至 2019-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=659421
• 关键词: Symbolic; dynamics; model; based; control

Autonomous driving technologies will transform the automotive sector, greatly increasing vehicle safety and transportation efficiency, and providing low-cost mobility to thousands unable to handle the task of driving on their own. Private and public sector investment is ramping up in the expectation of significant progress over the next few years. Google's ambitious program promises fully automated vehicles by 2020, but Google's very expensive state-of-the-art vehicles are achieving only 2000 km between forced disengagements of the driverless system.**A promising technology to control autonomous vehicles so that they avoid static objects and moving cars or pedestrians is known as "model-predictive control". Essentially, a dynamic model of the car predicts a short time into the future; the controller uses this prediction to control the steering, brakes, and throttle. The better the model, the better the control. However, the model-predictive controller must run very quickly on a car's computers, especially when emergency maneuvers are required.**Symbolic computing is a technology pioneered and commercialized by Maplesoft. It allows the direct manipulation of the mathematical equations within the model and controller. Not only can the equations can be simplified symbolically, they can be used to generate extremely fast simulation code - such as that needed by an autonomous vehicle.**The goal of this research is to develop new theories and symbolic computing software that automatically generates very fast model-predictive controllers; the target application is the University of Waterloo's AutonoMoose, the first autonomous vehicle approved for Ontario's roads. The proposed new theories and Maplesoft-based algorithms will improve the performance of vehicle controllers, and accelerate the development of autonomous vehicles in Canada and abroad. In turn, this will increase the adoption of Maplesoft software by automotive manufacturers and researchers around the world.****

自动驾驶技术将改变汽车行业，大大提高车辆安全性和运输效率，并为数千名无法自行驾驶的人提供低成本的机动性。 私营和公共部门的投资正在增加，预计未来几年将取得重大进展。 谷歌雄心勃勃的计划承诺到2020年实现全自动驾驶汽车，但谷歌非常昂贵的最先进的汽车在强制脱离无人驾驶系统之间只能行驶2000公里。一种有前途的控制自动驾驶汽车的技术，使它们避开静态物体和移动的汽车或行人，被称为“模型预测控制”。 本质上，汽车的动态模型预测未来的一小段时间;控制器使用此预测来控制转向，刹车和油门。 模型越好，控制就越好。 然而，模型预测控制器必须在汽车的计算机上运行得非常快，特别是在需要紧急操纵时。符号计算是由Maplesoft开创和商业化的技术。 它允许直接操纵模型和控制器内的数学方程。 这些方程不仅可以符号化地简化，还可以用于生成极快的仿真代码，例如自动驾驶汽车所需的代码。这项研究的目标是开发新的理论和符号计算软件，自动生成非常快速的模型预测控制器;目标应用程序是滑铁卢大学的AutonoMoose，这是第一辆被批准用于安大略道路的自动驾驶汽车。 本文提出的新理论和基于Maplesoft的算法将提高车辆控制器的性能，加速加拿大和国外自动驾驶汽车的发展。 反过来，这将增加全球汽车制造商和研究人员对Maplesoft软件的采用。