One important issue for the management of mobile actors in a wireless sensor and actor network (WSAN) deployed in a safety-critical environment is to route and schedule the mobile actors in an energy-efficient way to timely respond to the events captured by sensors. This problem can be modeled as a multiple traveling salesman problem with time deadlines. In this paper, a variable neighborhood search (VNS) approach is used to solve the routing problem. We compare two variants of the approach: one only accepts a better solution in the search and the other also accepts a worse solution probabilistically. Numerical experiments show that the first method can get better solutions whereas the second one is much faster. Both methods can obtain satisfactory solutions.