Design of an automated street crossing management module for a delivery robot

Autonomous navigation of mobile robots in urban environments is a complex problem, that can be decomposed in several tasks. Among them, autonomous street crossing is particularly difficult because it requires the robot to estimate the position and speed of surrounding vehicles and to decide which is the best action to perform based on such information. This paper develops the entire pipeline that implements autonomous street crossing; the approach is composed of an extended target tracking algorithm that estimates the position and velocity of obstacles and a crossing algorithm that determines the best crossing strategy to negotiate an unregulated intersection (i.e. without traffic lights) based on the other vehicles’ behavior. The method is first validated in an ad hoc simulation environment, and then experimentally tested using a prototype parcel delivery robot operating in a real urban environment. The results show that the robot is capable of tracking incoming vehicles and managing the crossing with good performance, in terms of the time taken to cross the road and of actions performed by the robot during the interaction with vehicles.