| Autor: |
Karsten Berns, Qazi Hamza Jan, Jan Markus Arnold Kleen |
| Rok vydání: |
2021 |
| Předmět: |
|
| Zdroj: |
Communications in Computer and Information Science ISBN: 9783030891695 |
| DOI: |
10.1007/978-3-030-89170-1_15 |
| Popis: |
Pedestrian zones are increasing in lengths and the relevant authorities are aiming at introducing autonomous vehicles to such zones for different purposes like transportation vehicles, delivery vehicles and cleaning vehicles. As the name suggests, pedestrian zones are crowded by pedestrian having highest priority. Vehicles in such zones should take special care regarding the safety of the people. Unlimited situations can occur from the pedestrians. Hence, it becomes crucial to test such vehicles before deploying in real-world to ensure safety of the people and the environment. Therefore, there is a need to have a realistic behaviors of pedestrians in simulation for critical testing of autonomous vehicles. The major novelty in this paper is the generation of virtual pedestrians which avoid obstacles realistically and to provide a platform for accurate and easy testing of autonomous vehicles in pedestrian zones. These virtual pedestrians avoid all kinds of obstacles based on self-awareness while progressing towards the goal. Main feature of these pedestrians are based on direction and speed heuristics. The goal is to have minimum number of parameters to generate test scenarios with realistic behaving pedestrians for autonomous systems. A thorough experimentation is done to validate the system. In the final experiment, a real-world scene is recreated from the simulation and the results are compared. It is seen visually that both the trajectories from the simulation and real-world highly resemble to each other. By slightly tweaking the parameters, one can generate different test scenario. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
