A combined hardware/software approach for fast kinematic processing

In this paper we present an integrated approach for very fast kinematic computations in multibody systems with complex structure. The approach relies on using optimized algorithms for generation of symbolic equations for kinematics together with application-specific hardware for their computation. T...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Microprocessors and microsystems 1998-09, Vol.22 (5), p.263-275
Hauptverfasser: Krieger, Ch, Hosticka, B., Krupp, Th, Hiller, M., Kecskeméthy, A.
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:In this paper we present an integrated approach for very fast kinematic computations in multibody systems with complex structure. The approach relies on using optimized algorithms for generation of symbolic equations for kinematics together with application-specific hardware for their computation. The kinematic equations are optimized by incorporating closed-form solutions wherever possible. For optimized hardware, a computer board has been developed, which includes a single CORDIC processor ( COordinate Rotation in DIgital Computer) for the computation of transcendent functions, as well as a DSP processor for implementation of basic arithmetic operations. To compute the kinematic equations a widespread method is to realize equivalent CORDIC processor arrays. Due to the large number of processors required, this method results in very high costs and is less flexible. With the board presented here it is feasible to emulate these CORDIC processor arrays with a single CORDIC processor, resulting in a good compromise between speed and cost and furthermore providing great flexibility for the kinematic computations of different multibody systems. To achieve a high computational speed, the delays resulting from the sequential re-use of the single CORDIC processor, different hardware and software approaches will be described.
ISSN:0141-9331
1872-9436
DOI:10.1016/S0141-9331(98)00057-X