||Wireless Sensor and Actuator Networks has become an important area of research, in
particular, due to their intrinsic characteristics, such as, flexibility, low operational and maintenance
costs and scalability. When dealing with systems’ supervision in industrial environments, Wireless
Sensor and Actuator Networks can be used to detect and classify transitory events. In this context,
it is essential that collected data should be reliable, ensuring the integrity of received information. A
particular case of sensed data degradation stems from readings outliers, which should not be sent to the
base station through the network. They must be locally detected and accommodated before transmission.
However, traditional outlier accommodation techniques cannot be directly applied to sensor nodes due
to computational and memory limitations. For this purpose, an outlier detection and accommodation
architecture relying on a multi-agent framework, is proposed in this paper. Results collected from a
laboratory testbed show the relevance of such approach.