Research On Mobile Robot SLAM Adapting To Search And Rescue Environments

Simultaneous localization and mapping(SLAM)of mobile robots has been widely applied to the normal area environments.The application of SLAM method in search and rescue(SAR)environment is of great significance to provide SAR team members with live maps of postdisaster sites,to mark the location of search and rescue trapped persons and sacrifices,to mark the characteristics of significant goals(such as exits,dangerous objects,obstacles,etc.).In particular,one of the major challenges is the simultaneous positioning and construction of maps in the context of disruptive loose structures and unknown environments.The positioning information and generated maps of SLAM process provide reliable and effective guidance for the search and rescue team members accessing to postdisaster site.Under the harsh conditions of search and rescue environment,the GPS signal is usually unreliable or even unavailable for robots in the process of constructing environment map,location,search and rescue area coverage,and exploration trajectory planning.The SAR SLAM faces the urgent requirements in performances of speedability,the object of interest(OOI)accuracy positioning,the full area coverage,object identification,and loop-closure revisited point detection.The researches on the SLAM method model establishment include the assistant perception of robot arm,the construction of self-organized mobile Ad hoc wireless sensor network(Ad-WSN)of SAR environment,the multi-robot and multi-sensor SLAM,the perception-driven hierarchical analysis,2D/3D map reconstruction,and the fusion information.The main works are as following:(1)Robotic arm-assisted detection model is set up to assist in identifying and locating OOIs such as the trapped,the sacrificed,dangerous objects,exits,doors,switches,etc.by touching,turning and grabbing the target objects.(2)Multi-robot deployment of WSN frame nodes.A series of relay nodes transmit the original data of robot detection and LRF measurements to the master robot for running SLAM algorithm.And simultaneously,the remote PC monitoring center surveils and manages the mobile robot SLAM process through the Ad-WSN access point.(3)Design of perception-driven exploration(PDE)hierarchical analysis algorithm model and program procedure.At the bottom of the planning level,grid maps can accurately describe obstacles in complex environments and irregular shapes.The middle standard metric layer contains multiple submaps,and each submap is used for local map processing.The top target layer integrates each of local sub-maps to generate a global map.(4)Three-dimensional map reconstruction of local target area.Based on the construction of 2D global map,integrated RGB-D three-dimensional map construction,6D-SLAM pose estimation,2D/3D modeling and local map reconstruction,combining ICPs algorithm,integration of visual features,optical flow information.Information of depth,dense vision,and 3D sense are employed to identify the trapped,sacrificed and other significant goals.(5)The fusion model of multi-sensor and multi-robot(MAM)information,measurement dataset and a variety of algorithms is established.By using the integrated probability particle algorithm and penetration navigation methods,the performances of SAR SLAM are improved.Through integrating data association and feature point classification algorithm,SLAM method based on Bayesian analytic learning of discriminant function,EKF and R-B particle filter,the robots identify the trapped,the sacrificed and significant targets(exit,dangerous objects,obstacles,etc.)at the scene of deformation and damage disasters,and simultaneously,the robots build environmental map and locate themselves.Using mobileRobot platform to control single and multiple robots,and to configure multiple sensors for simulations and experiments of SLAM algorithm with computer programming.The improvement of SAR SLAM methods,i.e.,the robustness of SLAM algorithm,and the information fusion effect of multi-sensor when GPS signals are discontinuous or missing in the unstructured postdisaster scenarios,has been verified.