Abstract: Abstract: This paper proposes a new corridor edge detection and navigation algorithm based on laser scanner data for greenhouse mobile robots. The type of laser scanner is SICK LMS-30206, and the data in each group include 361 distance values in a single scan. All the values will queue in an angle-ascending order, with 0 degree as the start and 180 degree as the stop. For the corridor edge detection task, the algorithm will intensify the fluctuating features of the range values in the same scan by background removing. Then the modified scan will be divided into 24 equal subsectors. The mean square error of range value of each subsector will be computed and compared. The subsector with the minimum mean square error of range value will be taken as the corridor candidate. And bi-direction expending search will be launched for the corridor candidate. The expending search will stop at the point whose adjacent range value difference is larger than a predefined threshold. When the search stops, the corresponding laser beam serial numbers of the corridor edges will be found. As other objects may be mistaken as corridor when they are located in the neighborhood area of corridor and share similar range values with corridor scan points, a pseudo-corridor removing strategy is also presented. With the help of a benchmark corridor width, the pseudo-corridor scan points can be removed. The corridor edges can be used to generate speed control orders for the mobile robot. A new developed front-wheel-driven mobile robot called Walle is employed as the experimental platform. Walle has 4 wheels. The 2 front wheels are equipped with servo motors. The 2 rear wheels are universal. It can change its heading direction in a differential walking mode. In the navigation process, the center of the scan line in the corridor is chosen as the navigation goal for each speed control sampling period, and the speed differential between the 2 driving wheels is the main control parameter. By analyzing the geometrical and kinematical relationships, a transfer function between the expected steering error and the speed differential can be built. Then the speed differential can be induced from the corridor edge information based on the transfer function. Finally, speed-altering control orders can be generated according to the speed differential. Here the new algorithm chooses to speed up one wheel and slow down the other for obtaining the expected speed differential. Experiments are conducted on Walle under different environmental conditions, including the indoor corridor, the outdoor curve routine, and the inner-greenhouse corridor. The robot can navigate itself in the indoor corridor scene with a mean horizontal error of 1.2707 cm and a mean square error of 2.6772 when following the corridor center line. Also, Walle can navigate itself in the outdoor curve routine and the inner-greenhouse corridor successfully. Because Walle lacks the self-localization sensors on board, the real-time position values can't be obtained and recorded for the outdoor curve routine and the inner-greenhouse corridor cases. Two website addresses of the experimental videos are given as remedial measure. The feasibility of the new algorithm is verified through the experiments mentioned above.