Establishment And Model Characterization Of Typical Operating Cycle Of Sweeper Based On Visual Perception

In recent years,while the people's living standards have been continuously improved,the amount of various types of domestic waste has also increased.Some of them are scattered on urban roads for various reasons,affecting the environment and sanitation,leading to an increase in the demand for domestic road sweepers.Traditional sweepers have environmental energy problems such as low energy efficiency and serious pollution.This is in contradiction with increasingly severe global pollutant emissions and energy shortages.At the same time,due to the cleanliness requirements and the impact of some driver operations,some sweeper drivers have often selected high-grade operations when they are operation on the road,causing waste of energy.This paper relies on the sub-project 3 of "Key Technology Research and Application of N2 / N3 Type Pure Electric Commercial Vehicle Power Platforms" in the National Key R & D Program New Energy Vehicle Key Project in 2018-"Intelligent Control of Vehicles and Performance Enhancement Technology ".This paper is a research on the characterization of typical operation cycle and scene models of cleaning vehicles using pure electric intelligent cleaning vehicles operated by intelligent sensing technology,which can provide an objective and representative evaluation of intelligent recognition algorithms and the embodiment of energy saving rates.In this paper,from August 2018 to August 2019,Beijing was selected as the collection location in central China,with four distinct seasons and road operations covering various climates.In this paper,by selecting 28 cleaning vehicles with the highest utilization rate,a total of 29,090,628 vehicle mounted mechanical cleaning equipment power data was collected.After that,in this paper,combining the power change and distribution characteristics of the cleaning system of the truck,it is proposed to convert the collected data into 4 standard values: 0kW(stop),28kW(low),39kW(normal),54kW(high).This paper analyzes the preprocessed data through short stroke definition similar to driving cycle,calculation of characteristic parameters,dimensionality reduction of principal component analysis,and k-means clustering algorithm to obtain 4 types of short stroke for jobs.Then this paper calculates the distance between all short strokes and the feature parameter vector of the class it belongs to,and selects 4 short working strokes that best meet the characteristics of the category.According to the proportion of various short strokes,it is scaled to form the typical operating cycle of the 1800 s sweeper.The low gear accounted for 79.67%,which was the main gear,the normal gear accounted for 17.67%,and the high gear accounted for 0.16%.After comparing with the overall data collected,the deviation value is ideal,which can reflect the performance of sweeping vehicles in China Typical operating cycle in some aspects.At the same time as the operation cycle study,this paper also carried out the work of power matching experiment in Chapter 3.In this paper,by constructing 1008 types of garbage distribution scenarios,using the project intelligent cleaning vehicle real-time perception scene information,manually selecting suitable gear operates,recording the intelligently perceived operating scenes and scenarios and manually selected gear positions as experimental data,and constructing them for training power data set that matches the model.The data is trained using a 3-layer neural network and a learning vector quantization algorithm in Chapter 5.After optimized comparison,the prototype vector trained by the latter is selected as power Matching model.This model can receive vehicle perception information and make gear decisions.Then,based on the power matching model,this paper uses a randomly generated scenario input model to judge,combined with the time of the typical operation cycle in Chapter 4,the speed of the sweeping truck,and other factors.This paper presents a method for characterizing a typical operation cycle model that converts typical operation cycle into several corresponding scenarios.It provides a reliable basis for setting up scenarios for evaluating smart sweeper indicators such as the energy saving rate and the recognition rate of the sensing system.