Waterloo Autonomous Golf Cart Testbed

自动驾驶汽车（AVs）提高了安全性、移动性、可达性、能源效率，并创造了巨大的经济增长机会。控制和运动规划方面的进步对自动驾驶汽车的部署至关重要，但主要挑战仍在继续，特别是在不可预测的、行人较多的环境（如游乐园和机场航站楼）和复杂的、多车的、高度动态的情况（如环形交叉路口和交叉路口）中使用自动驾驶汽车作为个人运输工具的复杂性，这些环境更有可能发生横向和纵向碰撞。

Automated vehicles (AVs) improve safety, mobility, accessibility, energy efficiency and create significant economic growth opportunities. Advances in control and motion-planning are crucial to AV deployment, but major challenges continue to arise, particularly related to the complexities of using AVs as personal transporters in unpredictable, pedestrian-rich environments (e.g., amusement parks and airport terminals) and in complex, multi-vehicle, and highly dynamic situations (e.g., roundabouts and intersections) where lateral and longitudinal collisions are more likely. Our team (Lashgarian Azad, Jeon, Dautenhahn, Fischmeister) is conducting cutting-edge research on safe, socially-compliant, and reliable AV control & motion planning for both pedestrian-rich and road environments. We are requesting a full-scale AV testbed, specifically, an autonomous golf cart (AGC), to test our new AV algorithms. The requested AGC will be used to conduct the required experimental work and data collections for validating and optimizing our methodologies for the following research topics: (i) traffic participant behavior predictions, (ii) socially-compliant motion planning in crowded environments, and (iii) safe decision-making and controls at roundabouts and intersections. We have developed AV controllers for roundabout driving that successfully integrate model predictive control and reinforcement learning to achieve safe motions in complex multi-vehicle traffic scenarios. We are also designing novel data-driven AV controllers with safety assurance using the concept of control barrier functions. Our early-stage research takes full advantage of virtual simulations and scaled car experiments, but we require a vehicle for full-scale experiments to validate our novel AV controls and decision-making techniques for practical situations. The requested AGC instrumented with on-board sensors and a programmable controller is immediately needed to serve as our full-scale testbed. The programmable controller allows quick deployments of revised versions of our AV algorithms and optimizing them for real-life situations. The requested testbed will enable unique opportunities for technological breakthroughs towards the safe and efficient use of AVs in pedestrian zones and complex, on-road scenarios. There are strong and growing demands for highly skilled people in the AV sector. The research enabled by the AGC will provide a unique training environment for the next generation of AV innovators, equipping them with hands-on experience in adopting motion planning and control technologies while advancing their knowledge in learning-based methods, human-machine interactions, and safety assurance techniques. In addition to driverless cars, self-driving delivery & on-demand mobility services, and autonomous shuttles, our research results will impact other industries like construction, mining, farming, and forestry, creating great opportunities to achieve major economic benefits.