Abstract: To improve planning efficiency, shorten path length and ensure a smooth operation of a hedgerow pruning manipulator, this paper proposes a rapidly-exploring random tree algorithm (a-bRRT*) based on gravity and target offset probability. The algorithm introduces the idea of target deflection probability and the gravitational force into the asymptotically optimal rapid exploration random tree RRT* algorithm, which can balance planning efficiency and path length. The comparison of the rapidly exploring random trees based on target offset probability (bias-RRT) with RRT* and a-bRRT* shows that, when the top surface of hedge is being pruned, the path length of a-bRRT* algorithm is 66.32% shorter than bias-RRT, and the planning time is 44.19% shorter than RRT*. When the side of the hedge is being trimmed, the path length of a-bRRT* algorithm is 67.17% shorter than bias-RRT, and the planning time is 73.87% shorter than RRT*. The simulation results show that the proposed a-bRRT* algorithm greatly improves the search efficiency and shortens the path length.