Multi-mode Household Cleaning Robot Garbage Classification And Prototype Design

With the development of social economy and the advancement of technology,cleaning robots have become an important assistant to reduce the burden of human labor in daily life.The commonly used cleaning robots are mainly divided into two categories:commercial large scenes(roads,shopping malls,airports,parks,etc.)and household ones that operate in the home environment.For home cleaning robots,the technology is relatively mature and there are already a large number of marketed products,the most successful of which is the automatic floor sweeping robots that have become massively popular.Nevertheless,in fact,there are still problems of low intelligence and single cleaning mode for home cleaning robots.Since there are many kinds of garbage in daily life,and the characteristics of different garbage vary greatly,and the usual home cleaning robots are flat in appearance and small in size,they usually only have wheel roll sweeping and wiping operation modes,so they can only sweep small objects such as dust and debris,which greatly limits their garbage handling capacity.The paper imitates the mechanism that humans usually adopt different operation modes according to the characteristics of the garbage itself when performing cleaning operations,and proposes a hierarchical garbage classification strategy based on course learning to make decisions on cleaning operation modes by obtaining garbage form,appearance attributes,and specific categories,and establish a simple multi-cleaning structure of household cleaning robot hardware platform,and design a prototype of the cleaning robot's operation structure,control system and other module,the paper mainly accomplished the following work:For the problem that the purposefulness of garbage classification lies in influencing the decision of subsequent operation mode,we developed the rules of hierarchical classification of garbage and the rules corresponding to the attributes of hierarchical classification and disposal mode,and produced the experimental data set of garbage.In order to be able to extract the categories of multiple layers of garbage,we propose a feedback network model-based course learning method,which starts from the coarse to fine thinking pattern of human cognition,considers the correlation between the attributes of multiple layers,and through the feedback deep network model based on course learning,we can learn the shape,appearance attributes,and specific categories of rubbish level by level.At the same time,in order to help the network focus on the learning of different levels of categories in different periods,we use the channel attention module to learn from the main Learn the role of different features on the concept of different levels of garbage,strengthen useful features and suppress less useful features,and complete the selective expression of features.Comparing with the direct classification of garbage,this model has good interpretability.The design of the overall structure of the indoor cleaning robot,the drawing of the3 D model of the operating mechanism and the modular assembly of the hardware object,finally built a physical platform of indoor environmental cleaning robot,which mainly includes the environment sensing system,control system and actuator system,through the robot physical development to fully explore the feasibility of this indoor multi operating mechanism cleaning robot.