Stereo vision-based algorithm for control of nonholonomic mobile robot

Abstract

U cilju ostvarivanja efikasnog i pouzdanog transportnog sistema, savremeni tehnološki sistemi se baziraju na upotrebi inteligentnih mobilnih robota za manipulaciju i unutrašnji transport materijala. Da bi se smanjila neodređenost u okviru dinamičkog tehnološkog okruženja, mobilni roboti su opremljeni sa stereo vizuelnim sistemom, pomoću koga pouzadno mogu da ocene udaljenost tehnoloških entiteta. U ovom radu, predstavljen je novi algoritam za stereo vizuelno upravljanje neholonomnog mobilnog robota. Glavni upravljački algoritam, zasnovan na računanju greške u parametrima slike (IBVS - Image based visual servoing), koristi se za tačno pozicioniranje mobilnog robota u željeni položaj. Da bi se izračunale greške u parametrima slike, koristi se algoritam za identifikaciju karakterisitčnih objekata na trenutnim i ciljnim slikama. Poređenje ovih algoritama izvršeno je na setu slika laboratorijskog modela tehnološkog okruženja, čija je akvizicija izvršena kamerama Basler acA1920-25uc. Na osnovu rezultata poređenja, KAZE algoritam za identifikaciju karakterističnih objekata je pokazao najbolje performance. Da bi se testirao i verifikovao rad stereo vizuelnog upravljačkog sistema, pored simulacije, izvršena su i dva eksperimenta na mobilnom robotu RAICO (Robot with Artificial Intelligence based COgnition) u laboratorijskom modelu tehnološkog okruženja. Eksperimentalni rezultati pokazuju efikasnost predloženog stereo vizuelnog upravljačkog sistema u ostvarivanju željenog položaja mobilnog robota, uz minimalnu ostvarenu grešku.

Requirements for an effective and reliable material transport system within advanced manufacturing environment can be fulfilled by using intelligent mobile robots to perform material handling and transportation tasks. In order to reduce the degree of ambiguity occurring in a dynamic manufacturing environment, mobile robots are equipped with a stereo vision system that can reliably estimate distance to manufacturing entities. In this paper, a new stereo vision-based algorithm for control of nonholonomic mobile robot is proposed. The main control algorithm, based on an error in image parameters (IBVS - Image based visual servoing), is used for positioning of a mobile robot in the desired location. For estimation of the error in image parameters, point features are extracted from the current and target camera view via feature detection and description algorithm. A comparison of these algorithms is made on a set of images obtained in laboratory model of the manufacturing environment by using Basler acA1920-25uc cameras. Based on the results of comparison, KAZE feature detection and description algorithm is proven to be best suited for this specific case. In order to verify the stereo visual control system, simulation and real-world experiments are performed. Two experiments are conducted on a mobile robot RAICO (Robot with Artificial Intelligence based COgnition) in a laboratory model of the manufacturing environment. Experimental results show the effectiveness of the proposed stereo visual control system and its applicability in reaching the desired location with minimal accuracy error.