Development of a MiL/HiL/AViL Approach to Pre-Deployment Testing of Low Speed Urban Road Autonomous Driving in the Context of the Smart Columbus Autonomous Shuttle Deployment Sites 2020-01-0706
Low speed autonomous shuttles emulating SAE Level L4 automated driving using human driver assisted autonomy have been operating in geo-fenced areas in several cities in the US and the rest of the world. These autonomous vehicles (AV) are operated by small to mid-sized technology companies that do not have the resources of automotive OEMs for carrying out exhaustive, comprehensive testing of their AV technology solutions before public road deployment. Yet, we have a large number of public road deployments of these AV shuttles including two deployment sites in Columbus through the Department of Transportation funded Smart City Challenge project named Smart Columbus. Due to the low speed of operation and hence not operating on roads containing highways, the base vehicles of these AV shuttles are not required to go through rigorous certification tests. The way these vehicles driver assisted AV technology is tested and allowed for public road deployment is continuously evolving but is not standardized and shows differences between different states where these vehicles operate. Safety of operation is achieved by first limiting the speed of operation to be below 25 mph, making sure that the driver takes over during difficult maneuvers, if necessary, like turning at intersections and using simple collision avoidance to stop automatically or through driver intervention every time an obstacle is encountered on the pre-determined path. Past experience with AV deployments in public roads has shown that while this low speed human assisted autonomy is a safe mode of operation, it is not capable of preventing accidents altogether as evidenced by several cases to the contrary including: un-attentive drivers not being able to take over control fast enough, the AV rear ended by other vehicles due to its driving pattern being unexpected for other road users, sudden emergency braking of the AV resulting in passengers being hurt, not being able to backup manually as an un-attentive truck backing into the road is definitely going to cause a collision and, yes, a pedestrian hitting the side of the AV shuttle. The AVs and AV shuttles deployed on public roads that are using these deployments for testing and improving their technology is not the right approach and safe and extensive testing in a lab and controlled test environment including Model-in-the-Loop (MiL), Hardware-in-the-Loop (HiL) and Autonomous-Vehicle-in-the-Loop (AViL) testing should be a prerequisite to public road deployment. While the larger technology companies and OEMs have the resources to do this, it is too costly for individual startup companies and also AV technology researchers to develop such tools independently. What is, therefore, proposed and presented for the case of AV deployments in Smart Columbus in this paper is the sue of open source and publicly shared and realistic simulation environments where such tests can take place in MiL and HiL testing. With this motivation in mind, this paper presents 3D modeling of a small preliminary, private AV test road around our lab, and the Scioto Mile and Linden AV shuttle deployment sites of Smart Columbus. Both the Unity engine and the Unreal engine environments were used for the modeling which can be used in the freely available simulators LG SVL and CARLA. The paper explains how the AV shuttle deployment sites considered were modeled using two different approaches. The rest of the paper concentrates on the Linden AV shuttle route and illustrates how the 3D environment is used to build 3D point cloud map combined with a vector map. To illustrate the potential use of such a simulator environment, an AV shuttle with a soft 3D lidar sensor is operated in the Linden route using NDT map matching for localization and an Autoware path following algorithm with pure pursuit steering to follow the AV shuttle path. The idea is that users will be able to easily implement and test their own AV functions in this environment. How the same approach can be extended to HiL and AViL simulations is presented in the paper along with how to add traffic, disturbances, uncertainties and faults to the overall system including the map for testing purposes.