No abstract

One of the problems faced by developers of artificial intelligence algorithms when creating car control systems is that the actions of other road users are difficult to predict and have a large variability. Even if we assume that all actions comply with traffic rules and participants do not make mistakes, that is, to bring the actual environment closer to the ideal, the task of automating vehicle control still contains many difficulties. This paper describes what difficulties exist in the field of predicting the trajectory of objects, shows concepts that will help in solving this problem, and also describes a particular method of forecasting, which allows you to make a forecast for cars moving along traffic lanes. The main forecasting stages and the results of testing the method collected by using a ready-made data set are given. The results presented in the form of a set of metrics, are compared with another algorithm for predicting trajectories. As a result, the advantages and disadvantages of the created solution were identified.

No abstract

The problem of creating a fully autonomous vehicle is one of the most urgent in the field of artificial intelligence. Many companies claim to sell such cars in certain working conditions. The task of interacting with other road users is to detect them, determine their physical properties, and predict their future states. The result of this prediction is the trajectory of road users’ movement for a given period of time in the near future. Based on such trajectories, the planning system determines the behavior of an autonomous-driving vehicle. This paper demonstrates a multi-agent method for determining the trajectories of road users, by means of a road map of the surrounding area, working with the use of convolutional neural networks. In addition, the input of the neural network gets an agent state vector containing additional information about the object. A number of experiments are conducted for the selected neural architecture in order to attract its modifications to the prediction result. The results are estimated using metrics showing the spatial deviation of the predicted trajectory. The method is trained using the nuscenes test dataset obtained from lgsvl-simulator.

No abstract