Controlling the longitudinal movement of heavy-duty vehicles based on optimal control can be a cost-efficient way of reducing their fuel consumption. Such controllers today mainly exist for highway driving, in which the velocity is allowed to deviate from a constant set-speed. For vehicles with varying velocity demands, for instance vehicles in distribution and mining applications, such controllers do not exist to the same extent. This paper describes an implementation of, and experiments with, an optimal controller in a real heavy-duty vehicle. The velocity profile of the driving cycle varies due to curvature and varying legal speed limits. These limitations are used together with road slope, actuator limitations, and driveability considerations as constraints in the optimal control problem. The problem is solved offline as a mixed integer quadratic program, which generates trajectories for the velocity and for freewheeling. These are used as reference for the existing cruise control functions in experiments in a Scania truck. Results in terms of fuel consumption and trip time are compared with a benchmark controller that mainly follows a fixed fraction of the maximum possible velocity. Solving the optimal control problem results in 18% reduction of the fuel consumption and 1% reduction of the trip time. Experiments with fuel measurements results in 16% reduction of the fuel consumption and 1% reduction of the trip time.