Design Of A Wheeled Robot Platform Based On Hardware/Software Coprocessing

The optimization of metrics, such as performance, flexibility, versatility and cost, is a problemembedded system development has to face. On the other hand, as an important developingdirection to solve traffic and industrial problems currently, intelligent mobile robots is attractingpeople’s attention. In order to solve the problem of optimization of metrics in developing wheeledrobot embedded system, a layered asymmetric multiprocessing platform is proposed in our design,and applied for a wheeled robot, with the Hadware/Software(HW/SW) co-processing to improvedesign efficiency.The system is divided into3layers: high-performance process layer, real-time control flowlayer and data plane layer. FPGA(Field Programmable Gate Array) in data plane layer is used forcomputationally intensive tasks, while ARM microcontroller equipped with RTOS is used forcomplex real-time control and application extension to the high performance layer. A lose couplechannel are adopted to make the communication effective and flexible between data plane layerand real-time control layer. The role of FPGA in data plane layer can be a pre-processor,post-processor and co-processor of the ARM microcontroller. Different ways, such as wirelessnetwork and UART(Universal Asynchronous Receive and Transmission), are adopted to connectthe real-time control layer and high-performance layer, which can be a PC or a smart phone.Autonomous navigation based on machine vision and remote control and monitor based on humanvison are popular technology to realize an intelligent mobile robot. Applying the proposed HW/SWco-processing architecture to an autonomous navigation robot based on machine vision andpartitioning its control system’s HW/SW, the FPGA is divided into image pre-processor, DC motorcontrol co-processor, LCM post-processor etc. User defined IP(Intelligent Property) in FPGA canbe easily attached to local bus and communicate with the CPU through a universal on-chipmechanism. Dynamic threshold, image thinning and compress are implemented in the image pre-processor. The A3principle is adopted in DC motor double closed loop control. Effectivedisplay on the dot matrix LCM is implemnted.The HW/SW co-processing architecture and co-design methodology tremendously boost thesystem performance and accelerate the design. The robot can not only autonomously drive alongthe line marker, but also be remotely controlled and transfer video through an Android smart phone.The mobile robot is just one application case of this platform which also can be applied in manyfields of embedded systems and industrial control area.