Enhanced Social Force Model for Service Robots in Various Environments: Calibration and Simulation

With the increasing emphasis on social-awareness navigation of service robots, traditional social force models require improvements and recalibration. This paper proposes an enhanced social force model that integrates robot and group forces with the traditional one. The model is simulated in various scenes of the JRDB dataset, which is categorized into three categories — outdoor, indoor&open, and indoor&narrow. We use the maximum likelihood estimation method and genetic algorithm to calibrate parameters separately for each category and provide an explanation of optimal parameter values based on environmental characteristics. Through comparisons of predicted trajectories with real ones, we demonstrate that our model enables robots to behave more like humans in different environments.