Architectures of modern autonomous robotic systems (RS), such as self-driving cars or autonomous drones, are highly complex, involving multiple components interacting with each other. It is thus essential to ensure that the developed software for autonomous RS is robust and enables these systems to withstand the numerous challenges arising in dynamic real-world environments. To improve the testing effectiveness, optimization techniques are typically used to generate test inputs for the system in a simulated environment. However, one of the big problems in applying optimization algorithms to test generation is finding a good representation of the problem. In our project, we propose using artificial intelligence algorithms to automatically infer an efficient representation and perform the optimization in the learned latent space.
- 4 ай бұрын
Dmytro Humeniuk - AI for the safety of autonomous robotic systems
- Рет қаралды 16
Пікірлер