A biologically motivated approach towards modular and robust low-level sensor fusion for application in agricultural machinery design

In this paper a low-level sensor fusion approach inspired by distributed decision making in swarms of social insects is proposed for application in agricultural machinery. In contrast to the state-of-the-art this approach is not dependent on a system or sensor model. Instead it rather employs a majority voting heuristic-based on the relative sensor distances. The ability to deal with very sparse system information and computational resources makes it ideally suited for on-board use in the agricultural domain. This is because it can be utilized to combine sensor data across machine or manufacturer borders. The most prominent and common example for this are tractor implement combinations. As shown in the simulations and experiments the fuzzy voter provides the ability to reliably identify poor measurements and eliminate conflicting data. This way a consistent data basis is provided that can be employed for higher level functions on the machines.

► Robust heuristic-based low level fusion for wide range of applications.
► Modular for use in reconfigurable agricultural machine combinations.
► Sensor monitoring/sensor failure detection.
► Self-adjusting to changing environmental conditions.
► Meets needs of embedded systems (low processor/memory demand).