The article presents a study on reward function design for crowd simulation via reinforcement learning. The authors argue that the design of the reward function is crucial for successful simulation of human crowds. Their work provides theoretical insights on certain reward functions and evaluates them empirically, using energy efficiency as the metric. The findings indicate that directly minimizing energy usage is a viable strategy when paired with an appropriately scaled guiding potential. The study contributes to the development of new crowd simulation techniques and the wider study of human-like navigation.
Publication date: 25 Sep 2023
Project Page: Not provided
Paper: https://arxiv.org/pdf/2309.12841
