Empowering road vehicles to learn parking situations based on optical sensor measurements

Abstract

Connected road vehicles are about to create a massive Internet of mobile things, equipped with increasing sensing capabilities and autonomy. In particular on-board distance sensors allow for detecting free road-side parking spaces when passing by. Upon receiving parking information, other vehicles may efficiently navigate to free slots leading to decreased parking space search times. Yet, in real road situations, sensed information may be misleading due to the mobility of the sensor, driving on multi-lane roads, and unknown semantics of sensed free spaces. In this demo, we present a drive-by parking space sensing system consisting of a LIDAR optical distance sensor and a GPS receiver connected to a Raspberry Pi. By applying machine learning, parking situations are estimated. We demonstrate the effectiveness of our solution in standard parking situations, in presence of obstacles, and when overtaking bicycles and cars on multi-lane roads.