With the rapid development of vehicular edge computing networks, the massive amount of real-time data processing raises high demands on the computing capacity and energy efficiency of vehicles. In order to improve computational efficiency and optimize resource utilization, collaborative computing has become an effective solution. In this paper, a delay-aware collaborative computing scheme in vehicular edge computing networks is proposed. Firstly, in order to minimize energy consumption and ensure real-time task performance, the problem of collaborative computing with delay constrained data processing is formulated. Then, the duration of contact between task vehicles and nearby vehicles with idle computing resources is analyzed, and pruning algorithm is proposed to compress the solution space of the candidate patterns. Furthermore, an asynchronous advantage actor-critic algorithm is proposed to solve the optimal offloading decision and resource management scheme. The experimental results show that compared with existing collaborative computing methods, the proposed scheme can significantly reduce vehicle energy consumption and latency while ensuring task completion rate.