The paper "Sensor Fusion in Certainty Grids for Mobile Robots" by H.P. Moravec discusses the use of Certainty Grids, a numerical representation of uncertain and incomplete sensor knowledge, in mobile robot control programs. The authors highlight the effectiveness of this approach in sensor fusion, motion planning, landmark identification, and other central problems. They propose a software framework for their new Uranus mobile robot that will maintain a probabilistic, geometric map of the robot's surroundings, using various sensors such as sonar, stereo vision, and proximity sensors. The certainty grid representation allows for incremental updates, modeling the fuzziness of readings and combining multiple measurements to produce sharper map features. The map can be used for path planning, location identification, terrain classification, and object recognition. The authors also extend the certainty grid to detect and track moving objects and demonstrate how even simple versions of the idea can enable robots to perform tasks that were previously out of reach, such as exploring a region and returning to its starting point despite disturbances and terrain changes. The introduction emphasizes the limitations of traditional models that neglect uncertainties and alternative interpretations, leading to brittleness in robot controllers. The authors share their experience with a numerical representation used in the Stanford Cart in 1979, which demonstrated error tolerance and robustness.The paper "Sensor Fusion in Certainty Grids for Mobile Robots" by H.P. Moravec discusses the use of Certainty Grids, a numerical representation of uncertain and incomplete sensor knowledge, in mobile robot control programs. The authors highlight the effectiveness of this approach in sensor fusion, motion planning, landmark identification, and other central problems. They propose a software framework for their new Uranus mobile robot that will maintain a probabilistic, geometric map of the robot's surroundings, using various sensors such as sonar, stereo vision, and proximity sensors. The certainty grid representation allows for incremental updates, modeling the fuzziness of readings and combining multiple measurements to produce sharper map features. The map can be used for path planning, location identification, terrain classification, and object recognition. The authors also extend the certainty grid to detect and track moving objects and demonstrate how even simple versions of the idea can enable robots to perform tasks that were previously out of reach, such as exploring a region and returning to its starting point despite disturbances and terrain changes. The introduction emphasizes the limitations of traditional models that neglect uncertainties and alternative interpretations, leading to brittleness in robot controllers. The authors share their experience with a numerical representation used in the Stanford Cart in 1979, which demonstrated error tolerance and robustness.