| With the development of science and technology and the arrival of Industry 4.0’s intelligent era,autonomous driving technology has become a hot topic and is beginning to enter our lives.At the same time,unmanned vehicles designed for indoor or enclosed areas are gradually maturing.However,due to the special nature of the application environment,this type of unmanned vehicle still encounters many problems.For example,indoor unmanned vehicles in a pre-positioning environment without GPS face higher uncertainty in motion environments than outdoor environments.There are also many more obstacles and the movement patterns of pedestrians are also full of uncertainty.In this case,it is necessary to design a higher precision,more stable,and more flexible fusion positioning system framework and navigation scheme for different indoor environments under this no-GPS situation.Therefore,this article conducted in-depth research on the high-precision positioning and navigation framework for indoor unmanned vehicles without GPS and achieved the following results.1.Analyzing the kinematic model of the relevant chassis used in this article,we completed the hardware analysis and selection,and designed and built a two-wheeled differential unmanned vehicle’s software and hardware system,and determined the overall system framework.Through extensive reading of relevant literature and information about the current status of unmanned vehicles,we determined that the focus of this research is on scenarios without GPS signal coverage or GPS signal blocking,such as hospitals,factories,offices,and enclosed parks.And for different application scenarios,the selection of unmanned vehicle chassis is mainly divided into two-wheeled differential types for indoor use and Ackerman motion models for enclosed parks or large factories,laying the hardware and mathematical foundation for the design of subsequent positioning algorithms.2.We proposed and designed a multi-sensor fusion positioning algorithm for two different scenarios: indoor and enclosed parks,and implemented the repositioning function of unmanned vehicles based on this.To solve the problem of low positioning accuracy of unmanned vehicles in indoor environments without GPS,this article combined the Extended Kalman Filter algorithm(EKF),the Monte Carlo positioning algorithm based on particle filters,and its improved version-the Adaptive Monte Carlo Positioning algorithm(AMCL)-to propose a multi-sensor fusion positioning framework for two-wheeled differential indoor unmanned vehicles with wheel odometer,inertial navigation unit(IMU),and single-line laser radar.For indoor unmanned vehicles with Ackerman chassis,a multi-sensor fusion positioning framework with wheel odometer,IMU,ultra-wideband module(UWB),and single-line laser radar was proposed,and the mathematical model of the sensors was introduced into the positioning framework for calculation.In order to solve the problem of particle divergence and failure to converge in the Adaptive Monte Carlo positioning algorithm when the unmanned vehicle just starts or after experiencing the kidnapping problem in SLAM,the above positioning algorithms were improved again to achieve the repositioning of unmanned vehicles without initial position or after losing position.The fusion positioning algorithm for two wheel differential unmanned vehicles has significantly lower accuracy than 0.1m after verification,and it has improved the accuracy by 63% compared to a single wheel odometer and 76% compared to a single laser odometer,achieving the expected effect.3.This paragraph describes the proposal of an improvement for global path planning,the design of an algorithm for multi-task continuous cruising of unmanned vehicles,and a system for multi-vehicle collaborative mapping and navigation based on ROS.To address the problem of insufficiently smooth paths in autonomous navigation of unmanned vehicles,the A*algorithm for global path planning is improved to solve this issue.To achieve the cruise function of unmanned vehicles between multiple task points,a multi-point navigation algorithm is designed.Finally,a multi-vehicle collaborative mapping and navigation solution is designed for large-scale scenarios to solve the problem of multi-vehicle fusion mapping and separate autonomous navigation.4.The multi-sensor fusion positioning algorithm,improved autonomous navigation algorithm,and multi-vehicle collaborative mapping and navigation scheme are applied to unmanned vehicles.The upper computer software part adopts Ubuntu18.04,and the robot operating system ROS Melodic is installed on it.The feasibility and accuracy of the proposed positioning algorithm are verified in indoor and simulation environments,and the improvement in accuracy compared to using only one sensor is evident.The feasibility of the proposed relocation algorithm is also validated.Finally,the improved A* algorithm for path planning in the navigation algorithm is verified through comparison with traditional A*algorithm in Matlab and unmanned vehicle simulation environments,and the function of multi-point continuous autonomous navigation is implemented.The ROS-based multi-vehicle collaborative mapping and navigation framework is validated on a mobile robot,which can achieve collaborative mapping and fusion of multiple indoor unmanned vehicles in a largescale environment,and subsequently realize synchronous navigation of multiple vehicles under the prior map. |
PDF Full Text Request