| 摘要 |
Each partial object of the unit, e.g. go from A to B, keep the uncertainty of your position below a given threshold, or draw up a map of the area and insert these landmarks into it, is allotted different bonuses and penalities. Performance weightings for the individual objects are found in connection with the need to perform them after an analysis of the bonuses and penalities and they are evaluated in a control unit. In the context of the process, a local planning horizon is given in which the environment of the unit is divided into grid cells. Preferred directions are stored for these grid cells which lead the unit to the shortest route to already known or unconfirmed landmarks in order to reduce the positional uncertainty or confirm a landmark. All the various routes possible within this grid are examined to find their contribution towards enabling the unit to reach the goal. Here, the sum of the costs and usefulness of each partial object along each route is formed. That route is chosen which results in the greatest usefulness or the smallest loss. The cyclical performance of the process finally results in a target lying outside the local planning horizon. |
