Correct registration of vehicle‐mounted sensors is a fundamental prerequisite for the perception capabilities required for automation. However, the use of a standard marker or artificial‐feature‐based approaches is often infeasible in environments that do not allow for additional infrastructure. This paper presents a method for sensor registration that overcomes this limitation by utilizing the geometric structure of the terrain surrounding the sensor platform. The method determines the information content of registration parameters in measurements of the terrain, and it updates only the subspace of those parameters with information. The performance of the method is demonstrated for registration of a sensor to a large mining haul‐truck and a swing‐loading excavator. The method is shown to successfully register the sensor to each vehicle using a surveyed topographic map of the terrain. Performing self‐registration using a map generated by the sensor itself is also demonstrated, however specific vehicle trajectory conditions are required to provide information on all registration parameters.

This paper addresses the problem of verifying a control system's knowledge about the shape and pose of an electric mining shovel's digging assembly. The need for such verification arises in order to ensure safe autonomous operation. The likelihood of unintended collision is reduced by confirming that the digging assembly occupies the region of space it is thought to occupy. We present two methods for verification. The first computes the probability that regions of key interest have the geometric form expected, subject to an allowed uncertainty, given measured point-cloud data. The second computes a likelihood distribution over a family of possible hypotheses by considering the level of support each range measurement provides to each hypothesis. The ideas presented extend, with appropriate adaptation, to other applications where it is necessary to verify the knowledge that a control system may possess about regions of space that are occupied from instant-to-instant. C