This paper presents a system which performs soil analysis and uses the acquired data to generate optimal step pattern and foot hold performances with respect to the terrain conditions. This approach also provides solutions for critical situations when the robot foot already slips or undergoes excessive sinkage. For soil analysis using the robot foot, a new sensor was developed and successfully implemented. This sensor is able to measure the relative movement between the foot and the ground. The soil properties are exploited using conventional control system optimization processes combined with fuzzy logic and neural networks. To serve as a test-bed for experiments, a half scale model of the walking machine RoboTRAC was constructed.>