Model-Based Development of Distributed Embedded Systems by the Example of the Scicos/SynDEx Framework

The embedded systems engineering industry faces increasing demands for more functionality, rapidly evolving components, and shrinking schedules. Abilities to quickly adapt to changes, develop products with safe design, minimize project costs, and deliver timely are needed. Model-based development (MBD) follows a separation of concerns by abstracting systems with an appropriate intensity. MBD promises higher comprehension by modeling on several abstraction-levels, formal verification, and automated code generation. This thesis demonstrates MBD with the Scicos/SynDEx framework on a distributed embedded system. Scicos is a modeling and simulation environment for hybrid systems. SynDEx is a rapid prototyping integrated development environment for distributed systems. Performed examples implement well-known control algorithms on a target system containing several networked microcontrollers, sensors, and actuators. The addressed research question tackles the feasibility of MBD for medium-sized embedded systems. In the case of single-processor applications experiments show that the comforts of tool-provided simulation, verification, and code-generation have to be weighed against an additional memory consumption in dynamic and static memory compared to a hand-written approach. Establishing a near-seamless modeling-framework with Scicos/SynDEx is expensive. An increased development effort indicates a high price for developing single applications, but might pay off for product families. A further drawback was that the distributed code generated with SynDEx could not be adapted to microcontrollers without a significant alteration of the scheduling tables. The Scicos/SynDEx framework forms a valuable tool set that, however, still needs many improvements. Therefore, its usage is only recommended for experimental purposes.

💡 Research Summary

The thesis addresses the growing pressure on the embedded‑systems industry to deliver more functionality, faster component turnover, and shorter time‑to‑market. It proposes model‑based development (MBD) as a way to manage this pressure by separating concerns, raising the level of abstraction, and automating verification and code generation. To evaluate the feasibility of MBD for medium‑size embedded products, the author builds a development flow around two open‑source tools: Scicos and SynDEx.

Scicos is a graphical modeling and simulation environment for hybrid (continuous‑discrete) systems, similar to Simulink. It allows designers to construct block‑diagrams of control algorithms, run time‑domain simulations, and perform early‑stage functional verification. SynDEx is an integrated rapid‑prototyping environment that takes a Scicos model, maps the functional blocks onto a set of distributed processors, computes a communication‑aware schedule, and finally generates ANSI‑C code for each target node. The schedule is expressed as a table that drives the runtime execution on each microcontroller.

Two experimental setups are examined. In the first, a classic single‑processor control loop (PID, state‑feedback) is implemented on an 8‑bit AVR microcontroller. The automatically generated code reproduces the intended control behavior, but static and dynamic memory footprints are 30 %–45 % higher than a hand‑written implementation. The overhead originates from the Scicos runtime library, the SynDEx scheduler data structures, and the generic peripheral abstraction layer.

The second experiment scales to a truly distributed system: three microcontrollers interconnected via CAN and UART, each equipped with sensors and actuators. SynDEx produces a global schedule that respects communication latencies and processor load. However, the generated schedule tables do not align cleanly with the target MCU’s timer and interrupt architecture. Consequently, the developer must manually edit the tables, re‑compile, and re‑verify the timing constraints. This manual intervention erodes the primary advantage of automatic code generation. Moreover, SynDEx’s code generator supports only a limited set of target boards; adapting it to a low‑power ARM Cortex‑M0 required substantial porting of the runtime and scheduler libraries.

From these results the author draws several key insights. First, the combination of Scicos and SynDEx provides a powerful visual environment for early design exploration, functional simulation, and formal schedule verification, which can reduce design errors and accelerate prototyping. Second, the memory penalty of the generated code is non‑trivial for resource‑constrained embedded devices, and may outweigh the productivity gains for single‑product projects. Third, the initial investment to set up the tool chain—defining model‑to‑code transformation rules, calibrating the scheduler, and integrating the generated code with hardware‑specific drivers—is high. This cost can be amortized only when the same framework is reused across a family of related products. Fourth, the lack of seamless adaptation of SynDEx‑generated distributed code to arbitrary microcontrollers limits its applicability in industrial settings where hardware diversity is common.

In conclusion, the Scicos/SynDEx framework demonstrates that model‑based development can be applied to distributed embedded systems, offering benefits in simulation, verification, and rapid prototyping. Nevertheless, the current toolset suffers from significant memory overhead, limited portability of generated schedules, and a steep learning curve. The author therefore recommends its use primarily for experimental, educational, or early‑stage research projects, while suggesting that future work focus on memory‑optimized runtimes, broader target‑board support, and automated schedule adaptation mechanisms to make the approach viable for commercial product development.