Since the wheels always slip a little, you need to measure and correct the robots position continuously. GPS is not fine enough and a gyro sensor does not provide the absolute position of your robot. Of course you could continuously measure all movements using gyro- and tilt sensors and aggregate these measurements to calculate the current position. However I assume that the limites accuracy of the sensors and rounding errors will produce a very bad overall result.
For example, lets assume you measure acceleration and direction 100 times per second and you move one second. Then you need to combine 100 values form the two sensors, wich will result in the addition of 200 small error, causing a great overall error.
In real live, autonomic robots usually measure their own position with a radar sensor, an optical sensor in combination with optical marks (e.g. IR lamps, or a black line on the floor, or some flashlights at the edges of a field), or some other marks on the floor (e.g. barcodes).