kindyngen is a toolchain to transform composable models of kinematic chains and queries thereupon into correct-by-construction solver code. Example solvers are forward position/velocity/acceleration kinematics or forward/inverse/hybrid dynamics. Additionally, the kindyngen toolchain enables to interleave those familiar algorithms with additional computations, including control or estimation that reuses the kinematic chain state as the most simple world model available to a robot.
Structurally, the kindyngen toolchain consists of two parts. First, the kindynsyn synthesizer to compute a kinematics or dynamics algorithm given a kinematic chain model. Second, the code generator that transforms the kindynsyn output to code. For now the target language for the code generator is C (supported by various software libraries).
Both, the synthesizer and the code generator are composable. On the one hand, they require the composable building blocks which we call steps (for the synthesizer) and fragments (for the code generator). On the other hand, a top-level context represents the overall composition. This composition includes the solver configuration (for the synthesizer) and an application template (for the code generator) that contains, for example, the main function.
In following tutorials we will provide a developer with the basic understanding of the kindyngen toolchain. This includes an overview of the terminology as well as the architecture. Afterwards we start with configuring solvers using the built-in functionality of kindyngen. Finally, the objective is to demonstrate how a developer can extend the toolchain with custom functionality. For this tutorial we focus on the specification of a Cartesian level controller and how it maps to the robot's joint-level hardware interface.