![\"\"](\"https:\/\/ubwp.buffalo.edu\/cavas\/wp-content\/uploads\/sites\/105\/2018\/11\/icave2_pic-1024x554.png\")
<\/p>\n<\/p>\n
Currently, there exist a few simulation tools\u00a0for AVs. One of the most advanced platforms is perhaps Waymo\u2019s Simulation Platform, which is reported to be able to simulate thousands of self-driving cars in a realistic virtual environment. However, the detailed working of the system is not publicly known and the platform is not available outside Waymo. Another interesting system is the NVIDIA DRIVE\u2122 Constellation AV simulator, which reportedly can simulate\u00a0the sensors of a self-driving vehicle\u00a0to generate photorealistic data streams to create a wide range of testing environments and scenarios. One of the limitations of this system is that it can test the autonomous vehicle systems built on the NVIDIA DRIVE\u2122 platform only. Another commercial solution is\u00a0<\/span>rFpro<\/span><\/i>\u00a0Driving Simulator. It can be used to simulate feeds to multiple sensors, such as cameras, radars,\u00a0and LiDAR. However, as most of the existing platforms, it is also not capable of simulating wireless communications, and connected and autonomous vehicle applications.\u00a0 There\u00a0are also a few open source simulation platforms. For example, the CARLA Simulator provides open-source code and protocols, open digital assets (urban layouts, buildings, vehicles) and two kind of sensors: cameras and simple Lidars, Unfortunately, it lacks abilities for traffic and network simulations. As mentioned earlier, none of the existing platforms and works has defined or developed any testing scenarios, performance metrics and scoring rubrics needed for benchmarking and standardization purposes.\u00a0<\/span> <\/p>\n
\nOne promising approach to testing, evaluating and certifying AVs is to leverage the unique\u00a0iCAVE\u00a0platform being developed at UB, which is an integrated\u00a0VR-based\u00a05-in-1 instrument, with both hardware- and human-in-the-loop. The\u00a0iCAVE\u00a0consists of: (1) Multiple driving simulators to conduct human\/hardware-in-the-loop experiments; (2) a microscopic agent-based\u00a0 traffic simulator with realistic traffic models and simulated driving environment; (3) a Network Simulator with realistic wireless communication models; (4) several instrumented vehicles, including a self-driving shuttle (the Olli bus)\u00a0and UB\u2019s Lincoln MKZ\u2019s AV platform\u00a0with sensor packages supporting both open source software called\u00a0Autoware\u00a0and Apollo; and (5) an instrumented environment with various traffic sensors and road-side units. Compared to road testing,\u00a0iCAVE\u00a0provides a safe and realistic environment enabling many \u201cwhat-if\u201d questions pertaining to futuristic and unproven AV technologies, and is also more flexible, has a lower cost, and supports larger-scale evaluation.\u00a0<\/span><\/p>\n