1,360
edits
Space Shuttle Navigation: Difference between revisions
Jump to navigation
Jump to search
→State vector
| Line 13: | Line 13: | ||
The numerical computation of the state vector can also be done if forces are acting, but the forces need to be known. [https://en.wikipedia.org/wiki/Inertial_measurement_unit <b>Inertial measurement units</b>] (IMUs), aboard the Shuttle, based on gyroscopes, sense any acceleration away from an inertial frame. Similarly functioning <b>Rate Gyro Assemblies</b> sense any changes in inertial attitude. | The numerical computation of the state vector can also be done if forces are acting, but the forces need to be known. [https://en.wikipedia.org/wiki/Inertial_measurement_unit <b>Inertial measurement units</b>] (IMUs), aboard the Shuttle, based on gyroscopes, sense any acceleration away from an inertial frame. Similarly functioning <b>Rate Gyro Assemblies</b> sense any changes in inertial attitude. | ||
The [[Space Shuttle Avionics]] hence knows where the Shuttle is at any time by computing the state vector numerically given orbital mechanics and the sensed forces. This procedure is error-prone, because neither are the equations of motion for the numerical prediction perfectly known, nor can they be computed to perfect accuracy, nor can the forces be sensed with perfect accuracy. Over time, the internal state vector known to GNC will therefore deviate from the true state vector. The purpose of the navigation systems is to supply extra sensor information to minimize the drift and to periodically correct the state vector. | The [[Space Shuttle Avionics]] hence knows where the Shuttle is at any time by computing the state vector numerically given orbital mechanics and the sensed forces. That's called the <b>propagated state vector</b>. This procedure is error-prone, because neither are the equations of motion for the numerical prediction perfectly known, nor can they be computed to perfect accuracy, nor can the forces be sensed with perfect accuracy. Over time, the internal state vector known to GNC will therefore deviate from the true state vector. The purpose of the navigation systems is to supply extra sensor information to minimize the drift and to periodically correct the state vector. | ||
Sensor information is always incorporated via [https://en.wikipedia.org/wiki/Kalman_filter <b>Kalman filtering</b>], i.e. GNC estimates a likelihood that any sensor reading is 'real' given the internal state vector and incorporates new data only if it fits into the overall picture. In this way, navigation does not go astray in case of sensor failures. Given a filter, any sensor reading can be characterized by a <b>residual</b> (difference between direct measurement and same quantity computed from the state vector) and a <b>ratio</b> (likelihood of the sensor reading being good given the overall picture). | Sensor information is always incorporated via [https://en.wikipedia.org/wiki/Kalman_filter <b>Kalman filtering</b>], i.e. GNC estimates a likelihood that any sensor reading is 'real' given the internal state vector and incorporates new data only if it fits into the overall picture. In this way, navigation does not go astray in case of sensor failures. Given a filter, any sensor reading can be characterized by a <b>residual</b> (difference between direct measurement and same quantity computed from the state vector) and a <b>ratio</b> (likelihood of the sensor reading being good given the overall picture). | ||
The basic characteristics of the navigational problem (including error propagation, sensor quality with respect to the propagated state, state vector corrections) are modeled in FG. | It's important to note that a small residual does <i>not</i> mean that the propagated state vector is close to reality - it just means that the propagated state vector is close to a sensor reading, but they may both be wrong the same way. In particular, there are state vector update procedures which allow you to replace the propagated by the filtered state vector. After executing such a command, the residuals will always be small, but the state vector will be wrong if the filtered state vector was bad. It's important to understand this and make good judgements. | ||
The basic characteristics of the navigational problem (including error propagation, sensor quality with respect to the propagated state, state vector corrections) are modeled in FG. However, it is possible to switch to perfect navigation in which the propagated state vector equals the true state vector using the simulation options. | |||
== Orbital Navigation == | == Orbital Navigation == | ||