Active stereo-vision 3D perception system for precise autonomous vehicle hitching

This paper presents a low-cost, low-power 3D optical sensor that is fully integrated for feedback control of an autonomous vehicle to a hitch. The hitching problem is an important challenge for vehicles in agricultural and transportation industries, yet few studies have been carried out in literature. We developed a perception system that combines low-accuracy classical stereo vision sensing with higher accuracy active sensing, as well as incorporating adaptive control strategies to tackle the hitching problem. We tested our proposed method on an RC car to approach a model hitch target. Our experiments demonstrated that our system can successfully control the vehicle from within ±35° of perpendicular to the hitch, to a final position with an overall standard deviation of 3.0 mm of lateral error and 1.5° of angular error.