Since 1981 the Mobile Robot Laboratory of The Robotics Institute of Carnegie Mellon University has conducted basic research on autonomous robots. We have built three mobile robots, and worked on two others; as a result of this work we have developed and improved new concepts in vision, sonar guided navigation, and robot control. These programs used a variety of approaches, some more successful than others. In the last three years we have concentrated on the most successful, a probabilistic representation of spatial knowledge we call Evidence Grids. Evidence Grids are a powerful and efficient unifying solution for interpretation of vision, sonar, laser and touch and other sensor data, for transforming sensorial information from multiple noisy sources into clear images of the robot’s surroundings. We think of them as providing a sense of spatial awareness for our robots.