Tham khảo tài liệu 'mobile robots - moving intelligence part 12', kỹ thuật - công nghệ, cơ khí - chế tạo máy phục vụ nhu cầu học tập, nghiên cứu và làm việc hiệu quả | Multi-Robot Systems and Distributed Intelligence The ETHNOS Approach to Heterogeneity 431 - Data filtering segmentation and aggregation activities are periodic or sporadic their period depends on sensor acquisition with a computational time at least one order of magnitude greater but still predictable . detection of color blobs building maps localization of important features etc. - Inter-robot communication is aperiodic with a non-predictable computational time in a distributed context a robot does not know in advance when and for how long it will receive messages from teammates. - Aperiodic computationally intensive activities can be optionally considered whose time cannot be easily predicted but - usually - is orders of magnitude greater collaborative self-localization symbolic planning etc. Notice that only the former four activities are critical for the system whereas the latter class includes activities that increase performance but are not critical. An overall architecture for the development of RoboCup players must take these aspects into account and thus permit the integration of less computationally intensive activities whose execution is critical for the system and therefore have hard real-time requirements with more computational intensive ones whose execution is not critical for the system and therefore have not real-time requirements . Periodic time bounded critical activities Sensor acquisition Reactive planning control Data filtering segmentation and aggregation are executed in high priority experts according to two possible algorithms Rate Monotonic or the non-preemptive Earliest Deadline First algorithm Jeffay et al. 1991 . EIEP allows exchanging information between real-time experts asynchronously thus avoiding unpredictable delays or priority inversion phenomena. Aperiodic not time-bounded non-critical activities are executed in low priority experts they run in the background and communicate their results to realtime experts whenever their .