Multisensor Fusion SLAM Research Based on Improved RBPF-SLAM Algorithm

Algorithm 2

The process of the improved RBPF-SLAM algorithm.

Step 1 Receive sensor data information and fuse sensor data from the odometer and IMU by the EKF method.

Step 2 Construct a robot motion model based on the fused positional information and estimate the initial positional at moment .

Step 3 Generate the map at moment based on the initial position of the particle. Then, search in the area near and calculate the match of observation with the existing map .

Step 4 When the scan match is low, the particle weights are updated directly based on the motion model sampling and calculation.

Step 5 When the presence of in the search region makes the match highly reliable, the observation model obtained from the LIDAR by scan matching is fused with the robot motion model to obtain a new proposed distribution function, and the particles are sampled from the improved proposed distribution.

Step 6 The parameters , and of the Gaussian distribution at the moment of the particle are calculated, and then the Gaussian distribution sampling is used to generate a new particle point set and update the particle weights.

Step 7 After updating all the particles at moment , the number of all valid particles is calculated to determine whether resampling is needed.

Step 8 Update the map for moment and reacquire sensor information for the next moment of map construction.