Technical Description
The input device consists of a circular platform built by two wooden discs 800 mm in diameter. Between both discs are three weight cells, arranged in an equilateral triangle. Silicon fat is used to enable the upper disc to slide along the tip of the weight cells, in order to minimise lateral forces on the weight cells which would falsify the measured weight shift.
The weight sensors can measure stretching forces as well as pressure. The sensors produce reasonable results within the weight range from 30 to 120 kg, assuming an average weight of 70 kg for a human navigator. Within the triangle built by the three weight cells, minimal shifts of the barycentric coordinates (ca. 5 mm, correspondin e.g. to moving an arm) can be detected. Outside this triangle the detection error may go up to 1.5 cm. The analog signal (ca. 10 per second) from each weight cell is enhanced by an amplifier and then forwarded to an analog to digital converter (ADC). The resulting digital signal is then transferred to a computer via a serial interface. A Personal Computer is used as an intermediate stage between input signal measurement (weight cells) and VR rendering (graphics workstation). The signals from the weight cells are sampled via three I/O ports. From these separate signals the barycenter coordinates of the navigator's body on the platform are calculated and sent to a graphics workstation over a fourth I/O channel for rendering of the new 3D scene. The reason for this, intrinsically unnecessary, intermediate stage is the instability of the (max.) 4 serial ports of the applied graphics hardware (Sillicon Graphics Onyx Re2).
Software: The data sent to the graphics workstation as input for scene rendering (based on the calculation of the navigator's position in a 3D model scenario) consist of two coordinates for the barycenter on the navigation platform, together with the total weight on the platform. From these data the back-transformation of the weight shift forces is calculated with the transformation equation mentioned above. The co- ordinates of the movement vector in the model scene (i.e. current 3D position, origin, direction, and speed) are updated accordingly. An application programming interface (API) library (based on the Polhemus Fastrak Library) is designed to enable the use of the information delivered by the basic functionalities of the Virtual Balance for a variety of application scenarios.