Trucks are largely used in road transportation worldwide in recent decades, and make great contributions to the GDP. Statistics shows that a large number of truck-involved accidents occur related to the ramp in merging manoeuvres on the highway. The accidents involved in trucks can lead to a considerable economic cost. The development of automated vehicl largely prompts the growth of vehicle active safety. In this thesis, we would like to address the problem of trajectory generation of trucks in merging manoeuvres. The trajectory generator is an important autonomous sector of a fully automated vehicle. A good generator can select the right moment for lane-changing manoeuvres by identifying a suitable gap in the traffic at the target lane, and define the path and acceleration profile. The potential algorithm, the optimisation-based algorithm and the sample-based algorithm are the main approaches to solve the motion planning problem. Due to its obvious drawbacks, the potential algorithm is not frequently used for the vehicle trajectory generation. The optimisation-based algorithm can get very accurate solutions; however, its performance depends on the development of the solver since most of the available solvers cost much when handling with nonlinear problems. The sample-based algorithm is a flexible method, and can be modified to fit a large range of situations. The sample-based algorithms are widely used in motion planning of small car-like robots. A lot of efforts have been paid on the studies of the sample-based algorithm. One of the most famous sample-based algorithms is the Rapidly-exploring Random Tree (RRT) algorithm. It is an algorithm that can cover the entire configuration space, and select a best path from the start state to the goal state quickly. However, the RRT algorithm is mostly used on small car-like robots with relatively low speed. Moreover, the results obtained by the RRT algorithm are usually not accurate. To address this problem, some modified RRT algorithms have been proposed. But most of the modified RRT algorithms have more complex structures, and achieve an accurate solution at the cost of more computational cost. The kinodynamic planning problem is to find a motion that goes from a start state to a goal state while satisfying all constraints of a nonlinear system. Trajectory generation of trucks is actually a kinodynamic planning problem. In this thesis, we would like to develop an RRT algorithm to solve the path planning problem of a large truck for merging manoeuvres on the highway. The basic RRT algorithm cannot solve the trajectory generation problem alone, since the basic RRT algorithm only focuses on the propagation of the tree and does not take the system’s dynamic into consideration. It should be incorporated with other algorithms to solve the path planning problem. The trajectory generated by the RRT algorithm cannot satisfy all constraints of the vehicle system. To solve this problem, a dynamic model that includes non-linear tyre properties with limits on longitudinal acceleration and steering angle, and with first order driveline and steering dynamics is incorporated in the RRT algorithm. Some strategies are incorporated in the RRT algorithm to improve the performance. The sample bias is introduced to the RRT algorithm to reduce the number of the waste samples by increasing the sampling probability of the nodes near to the goal region. The node selection method increases the choices of the nodes for the tree and reduces the sampling times. This method works by solving the nonlinear system for multiple times before adding the resulted node to the tree. The combination of criteria strategy uses different criteria to stop the algorithm when the goal is far away and is close. This method improves the efficiency of the algorithm. The RRT algorithm is developed for both online and offline implementation. The offline RRT algorithm is developed and implemented over the open-loop system and the closed-loop system. The simulation results show that a trajectory planned by the closed-loop RRT algorithm shows less disturbance and is smoother than a trajectory planned by the open-loop RRT algorithm. The real-time implementation is realised by updating the offline planning algorithm over closed-loop system. The replanning algorithm with regard to the highway situation is developed in this work. Collision avoidance is incorporated with the RRT algorithm by applying the intersection algorithm on the bounding volumes, since the intersection algorithm is a simple and quick method to detect the collision. The surrounding traffic is modelled as a set of Axis-aligned Bounding Boxes, and the truck is modelled as an Oriented Bounding Boxes in the configuration space. The intersection algorithm is applied on these bounding volumes to check the intersection of them. This work has developed a real-time RRT algorithm for trajectory generation for trucks for merging manoeuvres on the highway. We implement the RRT algorithm on both the open-loop system and the closed-loop system, propose ideas to improve the algorithm, and discuss the parameters that affect the algorithm in this work. It is a piece of work for the application of the RRT algorithm on the planning problems for large vehicles.