This paper introduces an optimal trajectory planner for the time-impact problem of redundant robotic manipulators, utilizing an enhanced grey wolf optimizer. Firstly, to address the limitations of the grey wolf optimizer (GWO) due to imbalanced exploitation and exploration, this paper introduces a reinforcement learning-based GWO (QLGWO) and its multi-objective version (MOQLGWO). In the QLGWO, Q-Learning guides search agents to choose exploration or exploitation actions based on experience and rewards, which is beneficial for achieving autonomous balance between local and global search. In the MOQLGWO, the archive and leadership selection mechanisms are used to search for Pareto optimal solutions that consider multiple optimization objectives. In addition, these mechanisms guide the search direction towards unexplored regions, enabling the algorithm to approach the global optimum. Secondly, two five-order polynomial functions are used to construct motion trajectories. The solution that needs to be searched consists of travelling time and joint positions, velocities, and accelerations at intermediate points. Finally, on 12 benchmark functions, the QLGWO was compared with the GWO and four other state-of-the-art meta-heuristic algorithms. Furthermore, the MOQLGWO was employed to solve the time-impact optimal trajectory planning problem for a nine-degree-of-freedom redundant robotic manipulator. Simulation and experimental results demonstrate that the QLGWO proposed in this study significantly enhances the performance of the GWO; the optimal trajectory planner can achieve a safe and smooth time-impact optimal trajectory, while adhering to joint constraints. The travel time of the searched trajectory is less than 14 seconds, and the impact is between -0.25 rad/s3 and 0.15 rad/s3.