Research On V-SLAM Navigation Technology Of Indoor Mobile Robot Based On Deep Learning

Visual simultaneous localization and mapping(V-SLAM)and path planning are the core technologies for indoor mobile robots to achieve autonomous navigation.Nowadays,with the development of deep learning,it provides a new research direction for the field of V-SLAM based mobile robots.At present,mobile robot autonomous navigation system still has problems,such as low accuracy of environment map and poor path planning performance.Therefore,it is great theoretical significance and practical application value to study the V-SLAM navigation technology for indoor mobile robots based on deep learning.First,vision sensors and deep learning framework are selected,V-SLAM and path planning technology are studied in this thesis.The overall design of V-SLAM navigation scheme for indoor mobile robots based on deep learning is completed.Secondly,closed-loop detection in V-SLAM systems is studied.For the problem that the mobile robot system has motion and estimation errors causing the inability to construct the same map.This thesis proposes an improved closed-loop detection algorithm based on convolutional neural network features.The method uses a pre-trained convolutional neural network(CNN)model as a feature extractor to extract shallow geometric features and deep semantic features of the input image from the network.The weight optimization of feature maps avoid the loss of spatial detail features.The similarity scores of the fused feature vectors are calculated for closed-loop detection.The experimental results show that the improved closed-loop detection algorithm based on convolutional neural network can effectively improve the image feature description,and the method has better accuracy and robustness.Thirdly,in path planning techniques,mobile robots in indoor complex environments using a single intelligent algorithm is prone to fall into local extremes,and multiple algorithm fusion leads to low time efficiency.So the improved particle swarm optimization(PSO)algorithm is proposed.The PSO algorithm is used as the main body,and the design environment selection strategy divides the particle swarm into two categories.The algorithm uses the selection,crossover,variation operators in genetic algorithm(GA)and the convergence operation in bacterial foraging optimization(BFO)algorithm to avoid local extremes.This scheme improves the planning efficiency of the robot.The experimental results show that the path planning based on the improved PSO algorithm is faster and shorter,it also has stronger robustness.Finally,the software platform is selected as Robot Operating System(ROS),and the hardware platform is built with the relevant equipment in the laboratory.The algorithms studied in this thesis are introduced to implement the designed autonomous navigation scheme for indoor mobile robots.The experiments demonstrate that the algorithm designed in this thesis can be implemented based on the platform.The V-SLAM navigation scheme of indoor mobile robot based on deep learning is real-time and reliable.