Open and Realistic smart City Activities Simulator (ORCAS)
| Title | Open and Realistic smart City Activities Simulator (ORCAS) |
|---|---|
| Summary | Create a simulation platform, loosely inspired by gamification, that is in principle capable of capturing the complexity of a complete city. |
| Keywords | |
| TimeFrame | Fall 2022 |
| References | |
| Prerequisites | |
| Author | |
| Supervisor | Sławomir Nowaczyk, Richard Bunk |
| Level | Master |
| Status | Open |
Create a simulation platform, loosely inspired by gamification, that is in principle capable of capturing the complexity of a complete city. It’ll start as a simulation of Halmstad University campus and grow from there. It’ll be open and designed to be endlessly expanded in collaboration with different partners, both in research projects and in education.
The main goal is providing data to train different kinds of AI and Data Mining models (including deep learning). There is a clear need for producing synthetic yet realistic sensor data in many areas. This is a common desire from many of our research partners. Data from such a simulation could be automatically labelled in a correct way, and thus cross-fed to other research areas at HU.
A unique simulation strategy (inspired by both movie- and gaming industries) will produce artificial sensor data greatly surpassing the quantity and quality of real data we have access to (e.g. through vehicle cameras) — both in temporal and graphical resolution. Together with an accurate physics engine, these simulations may, therefore, approximate reality to a limitless precision. The basic engines to provide this are readily available, what is lacking is an ecosystem of actors willing to provide realistic foundation -- such as infrastructure descriptions, behaviour models, etc.
The thesis can focus on one (or more) of the key directions:
- AI agent interaction. Find and implement rules that lead to self-org interactions/behaviour, e.g. commuting, consumption, etc.
- Sustainability: circular economy, reuse of resources. Examples include a mining industry (from ore to iPhone), governed by AI.
- Object generative algorithms (buildings, vehicles, humanoids, society). AI-driven, i.e., the generative algorithms learn design patterns/rules from examples.
- Simulation generated training data: synthetic and labelled, using the data for actual AI training (simulation ➾ data ➾ AI training).
Common aspects for all topics above include choice of the suitable graphic engine, eternal/infinite simulation (that can grow & improve during execution) and AI applications.
