Abstract: Abstract: This study aims to realize the real-time planning of obstacle avoidance paths for the agricultural machinery in the field operation using Bezier curve in the dynamic recognition area. The dynamic identification area of the agricultural machinery unit was firstly constructed for the operation walking, according to the tractor length and operation width, where the radar coordinate system (X-O'-Y) was established using the installation position of LiDAR. The obstacles were then identified in the dynamic identification area. The specific steps were as follows: 1) LiDAR was used to collect the point cloud data of the obstacle in the dynamic recognition area; 2) The point cloud information was processed into the three-dimension; 3) The obstacle was used to fit the obstacle envelope cylinder, then to project into the X-O'-Y coordinate system for the envelope circle (the center and radius). Then, the conversion model was established for the coordinate system, according to the installation position of Beidou Positioning Equipment and LiDAR, in order to realize the mutual conversion of longitude and latitude coordinates under the world standard longitude and latitude coordinate system (lon-o-lat coordinate system) and rectangular coordinates under X-O'-Y coordinates. According to the position of the obstacle in the X-O'-Y coordinate system, the obstacle avoidance path was divided into the execution and regression paths. When the ordinate of the circle center of the obstacle envelope contour circle was less than 1/3 of the length of the dynamic recognition area, the obstacle avoidance regression path was planned, otherwise, the obstacle avoidance execution path was planned. Finally, the obstacle avoidance path was planned as follows. 1) The coordinate system conversion model and Beidou positioning equipment were used to convert the collected current position of agricultural machinery from the longitude and latitude coordinates in lon-o-lat coordinate system to rectangular coordinates in X-O'-Y coordinate system; 2) The selection range of obstacle avoidance path control points using the third-order Bessel curve was calculated using the current position, heading angle, center coordinates and radius of obstacle envelope circle of agricultural machinery. The path cluster satisfied the minimum turning radius. 3) Taking the minimum curvature as the goal, the optimal obstacle avoidance path was selected from the path cluster for the coordinate transformation during tractor operation. The dynamic recognition area was reconstructed to repeat the above steps for the obstacle avoidance path. A real-time planning experiment on the obstacle avoidance path was carried out to verify the planning. The experimental results show that the maximum and average curvature of the obstacle avoidance path were 0.126 and 0.054 m-1, respectively, which were reduced by 25.9% and 42.6% than before, respectively. The maximum and average lateral error in the process of path tracking were 0.12 and 0.057 m, respectively, which were reduced by 36.8% and 28.8% than before, respectively. The distance from the tractor to the outer contour of the obstacle was greater than 0.375 m. Therefore, the obstacle avoidance path fully met the minimum turning radius, where the agricultural machinery can safely, effectively, and quickly avoid the static obstacles. The finding can provide technical support for the tractor unmanned operation.