Unmanned Aerial Vehicles (UAVs) have been deployed to enhance the network capacity and provide services to mobile users with and without infrastructure coverage. At the same time, due to the exponential growth of the internet of things devices (IoTDs), more and more data-oriented applications are coming up. However, as IoTDs have limited computation capacity and power, it is challenging to process collected data locally at the IoTDs. Motivated by the aforementioned facts, we propose, in this work, a UAV-assisted mobile edge computing system. Specifically, the objective of this work is to minimize the energy consumption of IoTDs, including local computation energy and uplink transmission energy, and UAV energy consumption. To achieve that, we formulate an optimization problem that optimizes the task offloading, bandwidth resource allocation, local computation resource allocation, and UAV computation resource allocation subject to the latency constraint of all IoTDs and the limitation of communication and computation capacity resources. Although the formulated problem is a non-convex problem, it is composed of convex subproblems, i.e., the formulated optimization problem in the form of a multi-convex optimization problem. Therefore, we decompose the formulated problem into convex subproblems and then alternately solve them till converge to the desired solution by using the Block Coordinate Descent (BCD) algorithm. Simulation results show that the proposed approach significantly saves the system power consumption compared to other existing schemes.