Speeding Up Robot Control Software Through Seamless Integration With FPGA - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2016

Speeding Up Robot Control Software Through Seamless Integration With FPGA

Résumé

Today robotic computing systems are usually implemented using general purpose processors because of their accessibility and simplicity which do not require specific knowledge. However, this approach restricts several optimization opportunities and may not always satisfy performance, cost, and energy requirements [5]. FPGA infrastructures can be considered as a good solution for these issues, especially in complex robotic systems that require time consuming tasks. Advantages are many to use FPGAs along with general purpose processors. On the one hand, FPGAs provide hardware acceleration and on the other hand, CPUs allow developers to use flexible software development environments. However, programming hardware devices requires a specific knowledge which remains a challenge for developers and usually results in a loss of productivity [3]. Moreover, interfacing FPGAs and high-level software remains problematic. This problem varies depending on projects due to the change of requirements and takes an important amount of development time. These difficulties are a barrier for a widespread adoption of FPGA in systems such as robotics ones. In this work, we present a generic architecture along with a toolset and libraries to ease the integration of FPGAs into existing robotic system. We provide a mean to connect the FPGA to the system and automatically deploy VHDL code that can be either legacy VHDL code, or generated VHDL code from a HLS tool. Developers do not have to worry about the interface between the FPGA and the processor. This interface is automatically generated according to the VHDL code. On the high-level software side, our software platform allows direct access to circuit registers on FPGA as memory addresses (memory mapping). To demonstrate this proposal, we describes a complete follower robot as a case study. The image processing to achieve a colour-based object detection is deployed on a FPGA. The rest of the robot controller runs on a CPU and has been developed using an object-oriented programming language (Pharo Smalltalk).
Fichier non déposé

Dates et versions

hal-01480221 , version 1 (01-03-2017)

Identifiants

  • HAL Id : hal-01480221 , version 1

Citer

Xuan Sang Le, Luc Fabresse, Jannik Laval, Jean-Christophe Le Lann, Loïc Lagadec. Speeding Up Robot Control Software Through Seamless Integration With FPGA. SHARC'16, Jun 2016, Brest, France. ⟨hal-01480221⟩
192 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More