We consider the scenario where a path-constrained ground vehicle is employed to collect data from the nodes that are massively deployed in a sensing field with varying importance levels in different regions. A novel data collection system model is presented by taking into account a region's important level. An optimization problem aiming at maximizing network utility and minimizing energy consumption under the vehicle's speed and path constraints is proposed. As the global network information is usually hard to obtain, a distributed algorithm requiring only the local information is derived. The proposed adaptive speed method works in an active manner and is able to satisfy the special constraints on data collection in the regions with high importance levels, which makes it more applicable than the constant speed methods.