Equivalência funcional de um sistema embarcado de código aberto para captura passiva e decodificação determinística de redes SAE J1939/ISOBUS em maquinário agrícola

Abstract

Contemporary Precision Agriculture is grounded in the interconnectivity of the elements that compose the agricultural ecosystem. However, the development of new technologies and data extraction are hindered by the high cost of equipment and the proprietary encapsulation of diagnostic tools. In light of this gap, this study aimed to develop and validate an open-source embedded system at up to 30% of the cost of commercial solutions for capturing, decoding, and recording structured data on CAN, ISOBUS (ISO 11783), and SAE J1939 buses. The physical architecture was implemented on a Single-Board Computer (Raspberry Pi 5) coupled with a galvanically isolated CAN transceiver, operating under the Linux networking subsystem. The decoding software was developed in Python, adopting asynchronous processing and concurrent thread structures for reading, decoding, and storing data contained in the Parameter Group Numbers (PGNs) of the SAE J1939 and ISOBUS (ISO 11783) standards. Validation occurred in two stages: initially through bench simulation via the injection of real logs extracted from agricultural machinery by third parties, and subsequently through a comparative dynamic field test, with the node connected to the diagnostic port of an agricultural tractor (John Deere 7230J), operating in parallel with a commercial PEAK PCAN-USB Pro analyzer. The bench results processed a sample window of 36,606 frames in 117 seconds with a leptokurtic profile (kurtosis 9.15), P99 < 2.20 ms, and a median latency of 0.44 ms. The field results confirmed the functionality of the proposed system compared to the commercial hardware, recording an identical sample window of 19,957 messages without the occurrence of packet drops. The interpreter maintained stable Central Processing Unit usage at 46.15%, with no evidence of thermal throttling during the proposed tests. The main limitations encountered were the restricted storage autonomy of up to 4.6 hours, the lack of thermal validation, and the existence of temporal anomalies of up to 928.86 ms due to Operating System preemption and the absence of a Real-Time Operating System (RTOS). It is concluded that the prototype enables the integral extraction of data from agricultural machinery networks, overcoming the cost and encapsulation barriers prevalent in the industry. By providing a validated open-source infrastructure for vehicular monitoring, the system positions itself as an instrument to foster access to precision agriculture, supporting the independent advancement of predictive agronomic modeling and Data Science within the digital agribusiness landscape.