Time-varying congestion and strict service time windows make pickup–delivery routing challenging because travel time depends on departure time. This thesis addresses the Time-Dependent Multiple Pickup and Delivery Problem with Time Windows (TD-MPDPTW), where each request includes multiple pickups followed by one delivery under precedence and vehicle capacity constraints. Time dependence is represented using time-of-day speed profiles and evaluated through forward time simulation, while the objective minimizes total travel time and waiting time from early arrivals. An exact Mixed Integer Linear Programming (MILP) is developed as a benchmark, and a Simulated Annealing heuristic is implemented for scalability. Experiments on synthetic instances up to 20 requests show that the MILP proves optimality for small cases (R=5) in about 25-30 s, while SA attains near-optimal solutions in <1 s with best gaps up to 3.1%, and remains feasible and competitive as instance size increases.