SAE/GM autodrive challenge
AutoDrive Challenge is a three-year competition among 8 North American universities to design, develop, and demonstrate a Level 4 automated vehicle capable of navigating an urban driving course.
Our "12th Unmanned" team took the second place in the first year competition in Yuma, AZ. The first year competition focused on vehicle development, sensor suite design, and driving within a static environment.
The perception team receives data from sensors such as camera, LiDAR, and radar. We use neural networks to detect the class and location of relevant objects. These classes include other vehicles, pedestrians, cyclist, drivable areas and traffic lights, traffic signs and more. We also use classical computer vision in order to find lanes lines. Our results are used by the path planning and decision making teams.
Path and Motion Planning
The path planning team provides guidance to keep the vehicle in the correct position at all times. The algorithm considers all the elements of the road and provide a safe path for the vehicle to take. This includes stopping for a pedestrian or avoiding an object (vehicle, bicycle, etc.).
The Vehicle Motion Control team ensures that the vehicle follows the desired trajectory as close as possible by implementing lateral and longitudinal controllers. The controllers output steering, braking, and throttle commands to the CANBus system.
The UIUX (User Interface & User Experience) team brings AutoDrive to everyday life. By software integration with ROS, we will present every detail of our masterpiece to the client. Although the AutoDrive Challenge hasn’t specified the need for a GUI (graphical user interface), no one wants to stare at dozens of command line windows, switching tabs and windows to understand the AV’s vision and decision. The UIUX exists to feed client information more efficiently.
The simulation team’s objective in the AutoDrive Challenge is to develop and test the Chevy Bolt’s controller’s response to oncoming vehicles and any obstacles blocking its lane as it travels in an urban environment. Specifically, the vehicle must be able to navigate around a block twice. In addition to navigating the city block, a delivery truck will appear and can perform a variety of actions that the Chevy Bolt must be prepared to handle. To meet this goal, the team will utilize Simulink in conjunction with an Unreal Engine simulation to test the vehicle’s controls.
The end goal of the Mechanical Team is to convert an existing SAE Level 1 vehicle into an SAE Level 4 vehicle. Our goal is to develop a sensor suite to support the perception activities.
Moreover, the mission of the Electronics Team is to maintain the distribution of power to all systems of the vehicle while maintaining the integration of the software. This integration must be without compromise and maintain full functionality throughout the vehicle. Lastly, our finished product must meet all SAE standards and prioritize the safety of the passenger above all else.
Year Two of the Mapping Challenge consists of utilizing provided offline HERE maps of North America and a GIS-mapping system to incorporate turn-by-turn directions, given an initial and terminal location. Both locations are to be easily identifiable, along with the best path visibly displayed on the map. Further, the system should offer routing and rerouting options, estimated travel times, and the ability to zoom in and out.