In UAV-assisted mobile edge computing (MEC) systems, the flexibility of unmanned aerial vehicles (UAVs) significantly extends service range and adaptability. However, UAVs are constrained by limited battery life and flight range, which restrict their ability to cover larger areas effectively. While some works consider vehicles as charging stations for UAVs, the potential for unmanned ground vehicles (UGVs) to collaborate with UAVs and serve as edge computing nodes remains largely unexplored. In this paper, we formulate an online joint task offloading and trajectory planning problem to minimize the total service delay in an MEC system with UAV-UGV collaboration, where UAVs not only land on UGVs for recharging but also use UGVs for long-distance transportation. The challenges of solving this problem include the long-term energy constraints, the time-coupled UAV-UGV collaboration decisions, and the non-convexity of the optimization problem. We propose a hierarchical solution framework to solve the problem. Specifically, we first employ Lyapunov optimization to handle energy constraints. We then use the sample average approximation method to handle the time-coupled UAV-UGV collaboration decisions, and utilize block coordinate descent combined with successive convex approximation to derive the task offloading and trajectory planning strategies. Theoretical analysis demonstrates the near-optimal performance of our proposed solution framework. Experimental results validate the effectiveness of our framework in reducing service delay.