Research On Improved Methods Of Piecewise Monte Carlo Localization

Robocup involves many fields such as artificial intelligence, image processing, andmulti-sensor data fusion. It was regarded as an important practice platform of relatedscientific and technological, as a result it called more and more attention by the scholars. Infootball matches, the localization is the primary problem to be solved for robots to completemany tasks, and great localization effect provides a firm protection for succeeding work.The environment of RoboCup is very complex, containing a lot of dynamic uncertainty,which makes self positioning difficult. Therefore, taking soccer robot in RoboCuphumanoid league as the researchbackground, this paper focuses on the study of Monte Carlolocalization method and ameliorates the performance of system by follows steps.1) This paper introducts piecewise passivity considerations for the particle degeneracyphenomenon. First of all, two states characteristic variables are defined. Secondly, the stateof system is determined according to the value of the characteristic variables. Lastly, thenumber of particles and the size of divergence factors are increased or decreasedcorrespondingly according to the regional state, and in the local tracking phase, newresampling rules are introducted to accelerate the convergence speed of particle.2) The kidnap problem is divided into big kidnap and small kidnap. A kind of statesreplacement method is proposed to sovle the big kidnap, leading the system come back toglobal localization phase. The size of divergence factors is increased to solve the smallkidnap, the system can also recover to the successful state of positioning after repairment.3) The Monte Carlo localization is improved based on negative information. While therobot is not seeing any landmarks that it expected to, the feedback information can bereturned to the positioning system, and the system will combine the negative informationand odometry information to update the particle set. So that the comprehensiveness andreliability of observation information of the robot is enhanced, and the positioningefficiency of the system is improved.Finally, to validate the improvement in system performance improvement, extensiveexperimental work is carried out, the experimental results show that the location of theimproved method is able to overcome the particle degradation problem and the robot kidnapproblem, the positioning efficiency and location precision is improved.