Research On Key Technologies Of Mobile Robot Teaching Platform For Higher Education

Mobile robot teaching platform is an integrated system which can achieve multiple information collection and interactions in a certain time and space.In more detail,it has the ability of environment mapping,localization,path planning,and meets the requirements of teaching and experiment standardization.With the development of the professional teaching system of artificial intelligence,the mobile robot teaching platform has become an important carrier of related majors.The existing robot teaching platform is derived from the Robot Development Platform(RDP)in the engineering field.It is generally matched with the short cycle,centralized and high-intensity training mode.It aims at the reproduction of experimental results and rapid technical mastery,and the theoretical principles behind it are less involved.While higher education emphasizes the systematization and long-term learning of subject knowledge.In addition to mastering the technology,the students also need to understand the principle behind the technology,and be able to research independently.Therefore,considering the need for higher education in artificial intelligence,this paper constructs a curriculum system about mobile robots for undergraduate and graduate teaching.Guided by the construction of the theoretical system and aiming at cultivating students’ ability of conducting research and independent innovation,this paper designs the mobile robot education platform considering key technologies of the mobile robot:perception,mapping,path planning and motion control.The following technical difficulties were found and solved during the construction of teaching content and experiment:1.When designing the teaching content of mobile robot environment perception,there are some problems,i.e.,environment information degradation of the monocular image,over-segmentation,and depth order reasoning error,which are usually brought by the low discriminative ability of feature captured by the previous methods.To this end,based on the 2.1D sketch,this paper proposes a distance measurement definition combining edge,entropy,compactness,and color terms and a weighted description operator,and realizes a monocular image depth order reasoning method to boost the feature representation to solve the problems mentioned above.2.When designing the teaching content of dynamic path planning based on the long-term perception of environmental elements and events with respect to time or space,there are some problems,i.e.,the misdetection of same-class distractors in the scene,and the loss of long-term tracking targets caused by occlusion or target moving out of the field of vision.For this reason,this paper explores the feature difference between the target,same-class distractors,and similar-appearance distractors,and proposes a Siamese-based network tracker based on multi-level feature weighting,constructs a feature fusion module,color constraint model,which realizes the fusion of image-level features and semantic-level features.The proposed tracker reduces the wrong detection of the sameclass or similar-appearance distractors.In addition,this paper addresses the target drift and the loss of target with the aid of the proposed global redetection mechanism.3.In the process of guiding undergraduate students to complete experiments for scientific researching,this paper proposes a softwarehardware cooperative architecture to solve the problem of cooperative adaptation between hardware devices and software algorithms.In terms of hardware structure,the mobile robot is divided into four parts:chassis and power mechanism,control unit,computing unit,and peripherals,which realizes the rapid assembly and dynamic expansion of the hardware platform.In terms of software architecture,the principle of "constraining and decoupling" is followed,and the paradigm of "platform framework=data structure+algorithm library+process engine+tool library+communication protocol" is taken to realize the decoupling of scientific research algorithm and conventional program.This paper also defines a multi-layer interaction logic for communication.The experimental results show the feasibility of this method in teaching and experiment of related courses.Combined with the above research,this paper implements a set of mobile robot hardware device and optimizes the hardware driver,key algorithms,and software structure.Finally,this paper establishes a unified technical standard,which provides a new solution for students’ theoretical learning and scientific experiment.