ACM SIGGRAPH Workshop: Computer Graphics for Autonomous Driving Applications
Organizers: Antonio M. López, José A. Iglesias-Guitián
Autonomous driving (AD) will likely be the core of future intelligent mobility. As recent events have demonstrated, autonomous driving already involves complex scientific-technical, ethical, and legal issues. The scientific-technical challenge is multidisciplinary as we are not only responsible for the development of the physical vehicles but also the sensors and the artificial intelligence (AI) on which they rely. One key question is how to assess the performance of AI drivers and ensure that the desired safety and reliability standards are reached. AI drivers require a variety of models (perception, control, decision making) that must be trained on millions of data-driven experiences. Assessing their performance requires, in part, an understanding of whether that “data” or raw information from sensors with associated ground truth conveying depth, motion, and semantics, is sufficient to cover the scenarios that will be encountered in operation.
In this context, Computer Graphics (CG) has emerged as a key field supporting both performance assessment and training of AI drivers. Latest advances in CG suggest that it is feasible to design the corner cases both for training and testing. Simulation allows us to drive millions of miles to assess the performance of AI drivers, as well as generate millions of episodes for training the models behind AI drivers. This simulation-based approach requires advances in procedural generation of realistic traffic infrastructure, realistic behavior of traffic participants (human drivers, cyclists, pedestrians), augmented and mixed reality for on-board videos, simulation of sensors (cameras, LIDAR, RADAR, etc.) and multi-sensor suites, automatic generation of accurate and diverse ground truth, and beyond real-time simulations on multi-AI agents.
The goal of this workshop is to bring together researchers and practitioners of both autonomous driving and computer graphics fields to discuss the open challenges that must be addressed in order to accelerate the deployment of safe and reliable autonomous vehicles. Speakers with experience on the use of simulation and computer graphics for autonomous driving will be invited to share their work and insights regarding upcoming research challenges.
- Automatic generation of accurate and diverse ground truth
- Modeling and simulation of sensors (cameras, LIDAR, RADAR, etc.)
- Procedural generation of realistic traffic infrastructure
- Realistic behavior of traffic participants (human drivers, cyclists, pedestrians)
- Real-time multi-agent simulation
- Augmented and mixed reality leveraging on-board data sequences (e.g. videos)
- Management of on-board large data streams
To participate in this workshop and the conversation please submit a CV and brief answers to the questions below. Responses should be limited to two pages. Follow this link to the submission system. The organizers will review the responses and curate a program for the workshop. A submission is required to be invited to attend the workshop as space is limited. There will be a nominal fee to cover lunch expenses, details for purchasing tickets will be forthcoming.
1.) Describe your current professional role and its relation to autonomous driving (AD).
2.) Briefly describe your motivation to attend this workshop.
3.) Please provide an example of specific questions or topics you would be interested in hearing or debating more about.
4.) Could you mention some tools involving computer graphics or driving simulators that you or your team often use or are aiming to use for your work in relation to AD? Feel free to comment about driving simulators that you might have used in the past and what you are looking for in the future.
5.) Please feel free to include any other comments you might want to share with us.
Applications will be accepted on first come/first served basis until July 27th. Apply to participate!