Mobile Platform Control Using a Neural Network Trained with Human Experience Data

The problem of mobile platform navigation has been extensively researched in the literature using a variety of methodologies and settings. To investigate the possibility of leveraging human skills in the fusion of sensor data, this study presents an artificial neural network-based method for guiding mobile platforms. It enables the control of a mobile platform while it moves between two locations in an environment where obstacles exist at different places. The target's position in relation to the platform and obstacle distances measured by sensors installed on it are the inputs used by the network to determine the appropriate steering to apply to the platform. By using data gathered from a human user relying on a visual interface and operating the platform in different rounds, the network is trained to acquire the logic of deciding how to steer the platform to avoid obstacles. The method is evaluated in a simulated setting. It showed that the neural network can effectively apply the knowledge it learned during training to control the platform's motion. The study shows that this type of learning, where the human user and the neural network do not rely on the same types of data to control the platform, can be used in mobile platform motion control, which is the paper's main goal.