Dynamic Task Assignment Framework for Mobile Crowdsensing with Deep Reinforcement Learning

Algorithm 3

MCS dynamic task assignment algorithm

Input: Task set L, worker set W, evaluation network Q

Output: Complete task set WL

 1: for from 1 to h do

 2: Obtain the worker set and the task set in the current period

 3: for from 1 to do

 4:  Obtain the feasible task set of worker in according to the

     spatiotemporal constraints of Equation (6) and Equation (7)

 5:  if is empty then

 6:   Workers stay in place or move a short distance at random in perception

      range, waiting for the next period to be allocated

 7:   Continue

 8:  end if

 9:  if Worker is the initial state then

 10:   Set the value of the Q network input of the worker as the virtual

      location label

 11:  end if

 12:  Generate state-action pairs for workers

 13:  Input the state-action pair to the Q network generated by Algorithm 1, and

     obtain the set of workers

 14: end for

 15: Generate worker task matching set according to Algorithm 2

 16: Update worker and task information

 17: Proceed to the next period's assignment

 18: end for