Driving safely requires multiple capabilities from human and intelligent agents, such as the generalizability to unseen environments, the decision making in complex multi-agent settings, and the safety awareness of the surrounding traffic. Despite the great success of reinforcement learning, most of the RL research studies each capability separately due to the lack of the integrated interactive environments. In this work, we develop a new driving simulation platform called MetaDrive for the study of generalizable reinforcement learning algorithms. MetaDrive is highly compositional, which can generate an infinite number of diverse driving scenarios from both the procedural generation and the real traffic data replay. Based on MetaDrive, we construct a variety of RL tasks and baselines in both single-agent and multi-agent settings, including benchmarking generalizability across unseen scenes, safe exploration, and learning multi-agent traffic. We opensource this simulator and maintain its development at https://github.com/ decisionforce/metadrive. * Quanyi Li and Zhenghao Peng contribute equally to this work.Preprint. Under review.

No abstract

No abstract

The correlation calculation model between landslide and mapping factors has a direct influence on the accuracy of landslide susceptibility mapping results. Using the Baihetan reservoir area as a case study, the effect of several correlation models on mapping landslide susceptibility is studied. The frequency ratio (FR) and the information value (IV) coupled BP neural network (BPNN) model was utilized to assess landslide susceptibility, with the mapping results of the single back propagation neural network (BPNN) model acting as a reference. The receiver operating characteristic (ROC) curve, the frequency ratio, and the susceptibility index distribution (mean value and standard deviation) are used to compare and assess landslide susceptibility values. The FR-BPNN coupling model is less precise than the IV-BPNN model. Findings from a single BPNN model for susceptibility mapping are less exact than those from a coupled model. Using the coupling model of the mapping factor correlation approach to assess landslide susceptibility has evident benefits, according to the study. The coupled model employing IV as the correlation method provides the most accurate and dependable susceptibility findings, and the mapping results are more consistent with the actual distribution of landslides in the study area. It can effectively direct disaster prevention efforts in the reservoir region.