Prototype autonomous vehicles are equipped with a wide variety of sensors to perceive the environment. Inputs from multiple optical, infrared, LIDAR and RADAR sensors are combined to help on-board computers identify elements in the environment and react accordingly. Such prototype vehicles are used extensively to develop and test new concepts for various technical problem areas, such as environment perception, localization and trajectory planning.When it comes to validating such concepts with the large number of situations they are likely to face in the real world, virtual/digital validation tools become necessary.
Participants in the challenge will compete to create simulated LIDAR point cloud data for any 3D object on the road. The generated point cloud should take into account the field of view of the LIDAR sensor and placement of the sensor on the vehicle. Input to the simulation will be provided in the form of an OpenDRIVE format file which specifies the road geometry and placement of elements on the road. A 3D model of the road will also be provided. The virtual road has multiple lanes, navigation markings, and various traffic signs along the route. For each position of the vehicle along the virtual road, the prototype must simulate the output data from the sensor at that point.
The simulation should perform a volumetric estimation of the objects on the road and generate LIDAR cloud points for each position of the vehicle along the road. The point cloud data output can be in JSON format, with spatial distribution of each point, and optionally the reflectivity of the surface.
1. Conceptual overview document explaining how the sensor is fitted on the car, the working principle and output.
2. Input and Output format specification, template code explaining calling and output conventions.
3. Virtual road 3D model
4. Virtual road specification in openDRIVE format
As part of their submissions, participants are expected to include the following:
1. A write-up explaining the approach they’ve used to solve the problem
2. An algorithm/pseudo code explaining the simulation steps
Selected participants will be invited to hackathon sessions at Bangalore which will happen in two(2) stages. The objective of the sessions is to create a working simulation based on the inputs provided.