A car-like indoor mobile robot is a kinematically constrained robot that can be modelled as a 2D object translating and rotating in the horizontal plane among well-defined obstacles. The kinematic constraints impose that the linear velocity of the robot point along its main axis (no sidewise motion is possible) and restrict the range of admissible values for the steering angle. In this paper we describe a fast path planner for such a robot. This planner is one to two orders of magnitude faster than previously implemented planners for the same type of robot. In addition, it has an anytime flavor that allows it to return a path in a short amount of time, and to improve that path through iterative optimization according to the amount of time that is devoted to path planning, The planner is essentially a combination of preexisting ideas. Its efficiency derives from the good match between these ideas and from various technical improvements brought to them.