Adaptive Multi–Robot Coverage of Curved Surfaces

Abstract

This paper presents two adaptive coverage algorithms for the deployment of multiple robots into discrete partitions over curved surfaces. The algorithms compute a metric tensor field locally on the surface in order to shape the robots' partitions in position, size, orientation and aspect ratio according to the present anisotropy. The coverage algorithms are further incorporated into a hybrid coverage method for the complete coverage of surface areas. Each robot iteratively deploys and adapts the partition, then subsequently sweeps its assigned area. The algorithms are demonstrated in simulations on di↵erent mesh models, including meshes reconstructed from real laser point cloud data.

Equitable persistent coverage of non-convex environments with graph-based planning

“…Over discrete representations and graphs, Voronoi-based load-balancing partitions are presented in Boardman et al (2016) and in Durham et al (2012), where it is combined with a gossip algorithm. Surfaces in 3D are treated in Breitenmoser et al (2010aBreitenmoser et al ( , 2014. Distributed vertex substitution is used in Yun and Rus (2014), with twohop communication to guarantee convergence to the locally optimal configuration.…”

Machining-Based Coverage Path Planning for Automated Structural Inspection

“…CPP for crawler devices has also recently attracted research interest in areas such as material handling and logistics [36], vacuum cleaners, agriculture [37], demining [27] and inspection [5,17,18,[33][34], Traditional path planning algorithms for such platforms has focussed on Configuration Space (C-Space) representations such as Voronoi diagram [38], regular & occupancy grids, generalised cones [39], quadtree [40] and vertex graphs. Additionally CPP for Aerial platforms has received attention for opportunities in surveillance [41], agriculture [42] and disaster and emergency management.…”

“…Previous research within the Centre for Ultrasonic Engineering (CUE) (University of Strathclyde) has focussed on mobile crawlers [4,9], localisation strategies and approaches through novel visual [10] and Bayesian methods [11], along with multi sensor (Ultrasonic, Visual and Magnetic Flux Leakage (MFL)) payload delivery [12][13][14]. Similarly inspired research has been undertaken internationally focussing on areas such as robot design [15][16], path planning for crawlers [5,[17][18] and component inspection [19][20]. All path-planning approaches described in [5,[17][18]] fundamentally design their method for optical or visual inspection deployment techniques, which coincidentally only perform surface inspection and do not perform internal imaging.…”

“…Similarly inspired research has been undertaken internationally focussing on areas such as robot design [15][16], path planning for crawlers [5,[17][18] and component inspection [19][20]. All path-planning approaches described in [5,[17][18]] fundamentally design their method for optical or visual inspection deployment techniques, which coincidentally only perform surface inspection and do not perform internal imaging. Internal imaging of structures, such as thickness mapping, is commonly undertaken using ultrasonic techniques which require consistent control of sensor-to-surface normality and typically feature a very much reduced singlepoint measurement inspection volume when compared to visual sensors.…”

“…Internal imaging of structures, such as thickness mapping, is commonly undertaken using ultrasonic techniques which require consistent control of sensor-to-surface normality and typically feature a very much reduced singlepoint measurement inspection volume when compared to visual sensors. The genetic algorithm [5], spanning-tree [17] and Voronoi [18] path planning algorithms presented do not therefore naturally directly translate to automated internal imaging and ultrasonic thickness mapping applications.…”