We are developing network-based applications for education, journalism, and entertainment where many users share control of a single physical resource. Our latest project, Sharecam, is a single robotic pan, tilt, and zoom digital camera. In this proect we explore algorithms for controlling the camera frame based on independent requests from online users. We propose a metric for the degree of satisfaction for each user and formulate frame selection as an optimization problem. We propose a computational geometry-based algorithm and its distrbuted version. For n users, the algorithm runs in O(n2) time. Its distributed versions run in O(n) time on the client side and in O(n log n) time on the server side.
Figure 1: Sharecam's collaborative camera control interface on the Internet. Each Internet-based user loads two image windows. The lower window is a fixed image of the camera's reachable range of view. Each user requests a camera frame by positioning a dashed rectangle in the lower window. Based on these requests, the algorithm computes an optimal camera frame (shown with solid rectangle), moves the camera accordingly, and displays the resulting live image in the upper window. Requests and camera frames are updated every 10 seconds.
1Graduate Student (non-EECS)
2Visiting Professor, Utrecht University, The Netherlands
Existing fixtures for holding sheet metal parts are generally bulky, part-specific, and designed by human trial-and-error. In this project, we propose unilateral fixtures, a new class of fixtures that addresses these limitations using modular fixturing elements that lie almost completely on one side of the part, maximizing access on the other side for welding, assembly, or inspection. The primary holding elements are cylindrical jaws with conical grooves that expand between pairs of part hole corners; each grooved jaw provides the equivalent of four point contacts and facilitates part alignment during loading. We present a two-phase algorithm for designing unilateral fixtures. The first phase assumes the part is rigid and uses 2D and 3D kinematic analysis of form-closure to identify all pairs of candidate jaw locations. For this analysis we propose and prove three new grasp properties for 2D and 3D grips at concave vertices, and a new quality metric based on the sensitivity of part orientation to infinitesimal relaxation of jaw position. The first phase also sets bounds on jaw cone angles. The second phase addresses part deformation with a finite element method (FEM) analysis that arranges secondary contacts at part edges and interior surfaces. For a given sheet-metal part, given as a 2D surface embedded in 3D with n concavities and m mesh nodes, the kinematic algorithm takes O(n2) time to compute a list of all unilateral fixtures ranked by quality, or a report that none exist for that part. The FEM deformation analysis arranges r secondary contacts considering m part elements in O(m(sup>3r). We have implemented both phases of the algorithm and report alignment data from experiments with two physical parts.
Figure 1: Unilateral fixture used to hold sheet metal parts
Figure 2: Addition of secondary jaws to minimize deformation as calculated using a FEM mesh
1Graduate Student (non-EECS)
2Professor, Ford Motor Co.
3Outside Adviser (non-EECS), Ford Motor Co.
4Outside Adviser (non-EECS), Ford Moror Co.
To facilitate training and planning for surgical procedures such as prostate brachytherapy, we are developing new models for needle insertion and radioactive seed implantation in soft tissues. We describe a new 2D dynamic FEM model based on a reduced set of scalar parameters such as needle friction, sharpness, and velocity, and a 7-phase insertion sequence where the FEM mesh is updated to maintain element boundaries along the needle shaft. The computational complexity of our model grows linearly with the number of elements in the mesh and achieves 24 frames per second for 1250 triangular elements on a 750 Mhz PC. We use the simulator to characterize the sensitivity of seed placement error to surgical and biological parameters. Results indicate that seed placement error is highly sensitive to surgeon-controlled parameters such as needle position, sharpness, and friction, and less sensitive to patient-specific parameters such as tissue stiffness and compressibility.
Figure 1: Simulation of needle insertion based on an ultrasound image of a human prostate cancer patient. Frame (a) outlines the prostate (in green) and the target implant location (small white dot) which is fixed in the world frame. Our simulation places a radioactive seed (large green disc) at this location (d). After needle extraction and tissue retraction, the placement error, the distance between the target and resulting seed location shown in (f), is 30% of the width of the prostate. Needle plans that compensate for tissue deformation can reduce placement errors like these that damage healthy tissue and fail to kill cancerous cells.