Improving mobile robot navigation performance using vision based SLAM and distributed filters

In this paper, we suggest a vision-based SLAM (simultaneous localization and mapping) method to improve navigation performance of mobile robot, which is used 2 encoders to calculate its position. If mobile robot is in building, tunnel or under ground facility, it is difficult to obtain navigation information from GPS only navigation system, because there are not enough visible GPS satellites. To overcome this limitation, DR (dead reckoning) system is required. However, as DR operation time goes by, the navigation error is increased because of accumulation of sensor error and noise. There are variety kinds of methods to reduce these errors. In this paper, we use a vision sensor and particle filter. Some clear points on vision sensor image are selected and tracked for error compensation. That is called a SLAM (simultaneous localization and mapping) method. In this paper, distributed particle filter is used to cope with nonlinear observation model and to deal with changing the number of measurements. Computer simulations are conducted to demonstrate the performance of suggested filter.