At present, high speed dual purpose rail/road vehicles employing fixed non-destructive testing (NDT) sensors are used to inspect rails. Due to the uncertainties in characterisation of the defects when they are detected at high speed, manual re-visiting of the defects by expert operators is required before any decision regarding track maintenance is made. This research has been driven by a desire from the rail industry for a robotic system performing faster than human operators and being capable to both detect and characterise rolling contact fatigue (RCF) cracks in rails with the aim of automating the existing manual inspection and enhancing its accuracy and reliability. This thesis combines expert systems technologies with robotic NDT to fulfil this aspiration. A great deal of effort has been spent to develop a robotic inspection trolley which can automatically detect and characterise the RCF cracks in rails using an alternating current field measurement (ACFM) sensor. It uses a rule based expert system (RBES) proposed to control the robotic trolley and more importantly process ACFM data for both detecting and sizing defects. The developed system can detect the possible presence of defects in railway tracks at high speed pass (5-20 km/h) and can automatically return to an identified defect location to perform a slower and more detailed scan (up to 20 mm/s) across a rail section to determine the size, depth and number of cracks present in that section.