Learning path planning (LPP), which aims to optimize the learning order of these involved learning materials, is a fundamental problem towards personalized education. Recent approaches formulate LPP into the sequence optimization that can be solved by a reinforcement learning (RL) algorithm. However, existing RL-based studies often ignore the prior knowledge and suffer from a high computation cost. To this end, this study proposes a prior-guided reinforcement learning (PGRL) model by designing a prior policy. By combining the prior policy into the guiding policy, PGRL implement a jump-start approach for LPP where the search cost can be reduced. The experiment result on two real-world datasets shows that the predefined prior policy not only decrease the computation cost but also improve the performance of LPP.