Solver Model

All AnimatLab models can be downloaded from this public GitHub repo:

GitHub - CAGoldsmith/BRDF-Solver: Source code for the inverse kinematic and inverse dynamic solver created for the development of Drosophibot IIGitHub

---

Figure 1: The AnimatLab models created for A) Drosophibot II and B) Scale-Drosophila

Figure 1A shows the AnimatLab model developed for Drosphibot II. Each leg segment and the thorax are modeled as rigid rectangular prisms. Each DoF is modeled as a hinge. Each mobile DoF was oriented to place its distal leg segment perpendicular to the proximal leg segment when possible. The limits for each DoF were then set such that the midpoint of the range would correspond to the average angle value found for the DoF during the animal kinematic analysis. DoF that were ultimately fixed in the robot model were similarly fixed at their average positions from the animal data.

Table 1: Leg segment lengths of Drosophibot II and a 140:1 scaled Drosophila. Drosophila segment lengths were averaged from fly 3D motion capture data then scaled such that the middle leg femur was 100mm. Dimensions from Drosophibot II marked with * were lengthened due to mechatronic constraints.

The robot’s leg segments were designed such that their relative proportions were similar to those of Drosophila with the length of the femur set to 10 cm. This constraint results in the robot being approximately 140:1 scale to the insect. Table 1 compares leg segment lengths of Drosophibot II with the scaled-up average dimensions of the Drosophila specimens recorded as part of the animal analysis in Section \ref{ss:proportionAnalysis}. Table 2 compares the dimensions between ThC joints (or where the ThC would be) on the thorax, with Figure 2 showing these dimensions for reference. In general, Drosophibot II’s leg segment lengths are within 16% of the scaled-up prototypical fly's. The lengths that are not within these margins are indicated by asterisks (*) in Table 1. In the case of the trochanter, the segment was lengthened to the minimum distance necessary for the bracket to rotate around the body of the chosen actuators, the Dynamixel smart servos. While this lengthening does produce significant differences from the trochanter lengths in the animal, it was unavoidable without a complete reconsideration of the robot's actuation.

Table 2: Major thorax dimensions between Drosophibot II and Drosophila. A visualization of these dimensions is provided in Figure 2

Figure 2: Top-down visualization of the major thorax dimensions from Table 2.

The thorax similarly had several dimensions changed due to the size of the actuators in the thorax imposed limits on the distance between the lateral pairs of legs and their anteroposterior distances, resulting in much larger errors between the robot and the scaled fly (Table 2). In order to keep proportionality across the entire thorax, each dimension therein was scaled according to the lateral distance between the middle leg pair ThC joints. Using this dimension as the ``constraining" dimension for the thorax produces the smallest viable thorax size for robot construction. The lateral distance between the middle leg pair ThC joints was lengthened from 10.00mm to 102.10mm because this was the minimum distance possible without the servos for the middle limb CTr joints colliding. The lateral distances between the front and hind limbs were then also extended by approx. 92mm to keep the lateral distances between the ThC pairs scaled similarly to the animal. Each leg was affixed to the thorax such that the CTr joints were approximately within the same horizontal plane.