Research On Indoor Multi-level Floor Plan Construction Based On Acoustic Ranging And Inertial Sensing

With the acceleration of urbanization and the increasingly complex internal structures of buildings,people’s indoor lives have become increasingly diverse,and location-based services has become an indispensable part of people’s daily lives.However,the main difficulty in popularizing indoor location-based services is the lack of accurate indoor floor plans.Dedicated mobile robots equipped with high-precision sensors can measure and produce accurate indoor maps,but their deployment rate is still very low due to high costs.The method based on smartphones uses computer vision technology to construct 3D point clouds,but it has some limitations,such as large image collection workload,privacy leakage,and susceptibility to lighting conditions.In order to address the drawbacks of existing methods,this thesis proposes a method of using smartphones to construct indoor floor plans.This method uses a combination of smartphone acoustic ranging and inertial tracking strategies to construct an indoor floor plan,and then refines the map to create a higher quality indoor floor plan.At the same time,this thesis also proposes an indoor navigation algorithm based on dynamic time warping,which can provide users with accurate indoor navigation.The main research work of this thesis is as follows:(1)In response to the complex and costly construction process of indoor floor plans,this thesis proposes an algorithm for constructing indoor multi-layer floor plans based on acoustic ranging and inertial sensing.This method emits frequency modulation continuous wave acoustic signals through the speaker of the smartphone,and uses microphones to receive echoes reflected from the wall.The distance between the smartphone and nearby walls is measured,and combined with the indoor reachable area generated by the user during walking to construct indoor floor plan.This method further refined the constructed indoor floor plan and demarcated the connection areas between floors,including stairs and elevators.(2)In response to the issue of accumulated errors in pedestrian dead reckoning,this thesis proposes a personalized step length prediction model based on LSTM.The model is trained by inertial measurement unit data with GPS data collected outdoors,and then infers the step length of users in the indoor walking process to achieve personalized pedestrian dead reckoning and improve the accuracy of pedestrian dead reckoning.(3)By utilizing inertial sensor data recorded on mobile phones,this thesis proposes an indoor navigation algorithm based on dynamic time warping.This algorithm records the inertial sensor data during the indoor map construction process,and uses the DTW algorithm to match it with the current inertial sensor data to achieve user positioning and generate accurate walking guidance.At the same time,the algorithm also includes a method of guiding the return of incorrect routes,providing users with more accurate indoor navigation.This thesis constructs a system prototype and conducts extensive validation experiments in teaching and residential buildings.Experiments have shown that compared with state-of-the-art methods,this system has better indoor floor plan construction performance,while addressing issues such as high construction costs,susceptibility to lighting conditions,and privacy risks.At the same time,indoor navigation experiments on multiple indoor scenes have also verified the effectiveness of the DTW based indoor navigation algorithm.