Bus driver scheduling deals with assigning drivers to vehicles which satisfy customers’ transport needs on tours. Safety concerns and labor laws dictate hard constraints like imposing a maximum driving time and taking required breaks. Bus operators seek for efficient schedules in terms of costs which should also be accepted by the drivers, for instance by avoiding duty parts cluttered over the day with long breaks and too many tedious vehicle changes. Decision makers want to learn about the trade-offs and the interaction of such conflicting goals. To this end, we compare different multi-objective optimization approaches to approximate the up to five-dimensional Pareto frontiers (PFs) for two real-world bus driver scheduling variants with complex sets of hard and soft constraints. We observe that a mutation-only NSGA-II already provides a solid baseline, while a population-based simulated annealing variant is often superior under short time limits when tuning it towards heavy restarting from the current PF approximation.