Underactuated tower crane lifting requires time-energy optimal trajectories for the trolley/slew operations and reduction of the unactuated swings resulting from the trolley/jib motion. In scenarios involving non-negligible hook mass or long rig-cable, the hook-payload unit exhibits double-pendulum behaviour, making the problem highly challenging. This article introduces an offline multi-objective anti-swing trajectory planning module for a Computer-Aided Lift Planning (CALP) system of autonomous double-pendulum tower cranes, addressing all the transient state constraints. A set of auxiliary outputs are selected by methodically analyzing the payload swing dynamics and are used to prove the differential flatness property of the crane operations. The flat outputs are parameterized via suitable B\'{e}zier curves to formulate the multi-objective trajectory optimization problems in the flat output space. A novel multi-objective evolutionary algorithm called Collective Oppositional Generalized Differential Evolution 3 (CO-GDE3) is employed as the optimizer. To obtain faster convergence and better consistency in getting a wide range of good solutions, a new population initialization strategy is integrated into the conventional GDE3. The computationally efficient initialization method incorporates various concepts of computational opposition. Statistical comparisons based on trolley and slew operations verify the superiority of convergence and reliability of CO-GDE3 over the standard GDE3. Trolley and slew operations of a collision-free lifting path computed via the path planner of the CALP system are selected for a simulation study. The simulated trajectories demonstrate that the proposed planner can produce time-energy optimal solutions, keeping all the state variables within their respective limits and restricting the hook and payload swings.