|
| |
The Virtual Football Trainer
From: University of Michigan
| By:
Klaus-Peter Beier |
EDITOR'S INTRODUCTION |
Engineers and scientists have been exploring the potential applications that cutting-edge virtual environments now offer the sciences, such as in accident simulation, medical imaging, architecture and urban planning, archeology, and industrial design.  What happened when Lloyd Carr, head football coach at the University of Michigan, tackled Klaus-Peter Beier (right), director of the University of Michigan's virtual reality laboratory, to build a Virtual Football Trainer? How well would the university's virtual "CAVE" environment, unrivaled in its ability to analyze spatial arrangements as found in complex mechanical systems, hold up against the brawn of the University of Michigan's football team? |
 magine watching the University of Michigan football team playing in a sold out Michigan Stadium on a sunny Saturday afternoon. But instead of cheering from the bleachers, you are transported right down to the field, in the middle of the action. You can move to any position and experience the game from the player's point of view. No longer a spectator, you feel like a participant in the game. |
The rapid development of virtual reality technologies makes this scenario possible, and opens up opportunities for new directions in team-sport training by investigating still unexplored areas of virtual simulations. While video games allow you to explore a football play by looking at a computer's monitor, immersive virtual reality provides a much different and unrivaled experience: You are on the field surrounded by virtual players, presented in full scale and in stereo. It seems that you can touch them. You can look and walk around, hover over the quarterback, or even fly to cover distances quickly. |
The Virtual Football Trainer allows coaches and players to study simulated formations and play variations from any viewpoint and as often as necessary. Training in immersive virtual reality can focus on the visual perception of a specific play situation, the correct estimation of distances, the awareness of the location of players on either team, the recognition of individual players (by size, stature, and player-specific movements), the visual communication with the coach on the sideline, the exchange of players on either team, the signaling by the referee, the changing display on stadium billboards, or the familiarization with a specific stadium through a simulated walk from the tunnel onto the field. |
All these exercises can be conducted in a controlled environment and can be repeated with high precision, as often as needed, and in a short period of time. No physical stadium is necessary and no additional players are required to practice a specific play. Even unpredictable weather is of no concern. The virtual players consistently behave as they have been programmed, are always there when needed, and are never tired. |
Flexibility in the use of the Virtual Football Trainer allows for the study of play animations outside the immersive CAVE system. Play animations can be distributed over the Internet or via CDs for visualization on a laptop or desktop computer. In addition, large screen projection systems can be used for the discussion of simulated plays during team meetings. |
Training football players in the virtual reality CAVE
The CAVE (Cave Automatic Virtual Environment) is  currently the most advanced system for immersive virtual reality. A CAVE provides its users with the convincing illusion of being fully immersed in a three-dimensional world that is computer-generated and presented to the viewer in realistic full scale as well as in stereo. |
A CAVE is a room-sized cube (typically 10 x 10 x 10 feet) consisting of three walls and a floor. These four surfaces serve as projection screens for computer generated stereo images. The projectors are located outside the CAVE. The users entering the CAVE wear lightweight LCD shutter  glasses that provide three-dimensional stereo viewing. The effect is so compelling that the walls and corners of the CAVE are mentally blocked out by the human brain. The floor projection allows three-dimensional objects (like virtual football players) to appear inside the CAVE room and confront the user in a convincing way. |
Through the surrounding walls, the CAVE provides ultimate immersion and an extreme wide field of view. Peripheral vision is well supported and instrumental for orientation, navigation and, most of all, for the perception of movements that occur in the margins. |
These aspects of the CAVE are instrumental for the Virtual Football Trainer. The objective is to provide players with the experience of visualizing play situations and the fast reactions of players on either teams. During the development phase of the Virtual Football Trainer, experts in the sport provided many scenarios that would be ideal for inclusion. Tom Brady, former Michigan quarterback, summarized some of the major applications after experiencing the Virtual Football Trainer first hand: - Young players can go through countless repetitions of their own team's plays and learn the hundreds of formations, motions and variations to an offensive and defensive scheme.
- An opponent's tendencies can be programmed and the simulation can be used to prepare quarterbacks and linebackers for anticipated schemes in an upcoming game.
- Players could spend additional time in the CAVE to review simulated defensive maneuvers, such as opposing blitzes versus pass protections, over and over again.
- Coaches could use the Virtual Football Trainer as a priceless tool in the off-season to train players. They could also visualize new plays to install in their own game plans.
|
The science behind the Virtual Football Trainer
The development of the Virtual Football Trainer required the application of scientific principles from a variety of disciplines including computer science, computer graphics, geometric modeling, human motion studies, character animation, artificial intelligence, user interface design, as well as human factors and human perception. As typical for any virtual reality application, the development team must be familiar with all of these areas and must apply an interdisciplinary approach to a specific application, in this case American football. |
 At the core of the application is the so-called hierarchical scene graph, a data structure that describes all elements of the virtual environment and their relation to each other. These elements range from the stadium and the field all the way down to a yard line or a specific skeleton joint of an individual player. A large number of computational algorithms were developed to create and to manipulate the scene graph and to interactively control the application. Such algorithms include, for example, collision detection, automatic character animation via a set of predefined rules, or time-controlled transform operations (scale, rotate, translate) applied to a player's skeleton. |
A virtual environment is always an approximation of the real world. To run the application in real time, simplifications are necessary. The knowledge of human perception principles allows for the creation of simplified environments that are still convincing. This is reflected, for example, in the design of the player's skeleton (which includes only the essential joints), the choice of colors and reflection properties for the various elements of the scene graph, the use of textures, or the illumination of the environment by virtual light sources. |
The rendering of the scene graph and the stereoscopic display in a CAVE system are based on 3D-viewing algorithms that are well understood and available via commercial software packages. |
Modeling and editing a play
 The creation of a virtual play begins with the design of a play chart entered in two dimensions using the "Chart Editor" program on a laptop computer. The players are presented by symbols, similar to the Xs and Os found in conventional playbooks. For each player, the initial position and the moving path on the field are defined via interactive graphics operations (using a mouse and pull-down menus). The play can be animated in two dimensions and the movements can be properly adjusted with respect to a time axis at the bottom of the screen. |
The Chart Editor supports this process with a variety of useful functions including a library of pre-defined initial formations for all 22 players and a collection of pre-defined plays that can be loaded and modified. Players can be added or removed and properties, such as a player's position, number and team (offense or defense), can be specified. The passing of the ball as well as the ball's flying path can be added and visualized. The finalized play can be labeled and added to a library. |
Creation of three-dimensional play animation
This most powerful component of the Virtual Football Trainer is fully automatic and uses artificial intelligence algorithms that create the three-dimensional animation for use in the CAVE. A three-dimensional virtual player, dressed in a team specific uniform and labeled with the player's number, replaces each two-dimensional symbol. The movements of these virtual players are created on a dense time grid. The algorithm recognizes if a player is walking, running, stumbling, tackling, and throwing or catching the ball. Using collision-detection methods, interactions between players that come into contact with one other are realistically modeled and portrayed. |
 To obtain realistic movements, each player is animated by means of an internal skeleton consisting of about 20 joints. A library of hundreds of pre-defined poses and movements defines the positions and angles for these joints on a time grid. For any given speed of motion, pre-defined movements are automatically fine-tuned and smooth transitions between poses and movements are inserted. |
The animated movement of a virtual player is controlled via an internal skeleton, a hierarchical structure of joints and links derived from human anatomy. The links are assumed to be rigid elements corresponding to the bones. The joints (marked by yellow spheres in the images, above and below) are the connecting points between the links and act like ball-and-socket connectors that allow for rotation at each joint. |
The skeleton is enveloped by a three-dimensional geometric shell that represents the player's external shape. This geometry is divided into segments with each segment corresponding to a specific link of the skeleton. A segment is always in a fixed relation to its corresponding link, i.e., if the orientation of a link changes, so does the corresponding segment. Ultimately, only the external geometry is rendered. The skeleton is never displayed, but is numerically embedded in the algorithm that controls the animation. |
Most joints have three degrees of freedom (rotation about the x, y and z axis, respectively). The challenge of a realistic animation is to determine the change of these angles on a dense time grid for all joints of the skeleton. With all 22 players on the field and each player moving differently and independently, a total of about 1,200 degrees of freedom need to be controlled simultaneously for the duration of an animated play. |
Controlling the play animation in the CAVE
The completed play animation is captured in a "Play-Script" file and transferred to the CAVE. The Virtual Football Trainer loads a given virtual environment (e.g., the Michigan Stadium) from a library and places the animated play at the proper location. |
In a typical training session, the trainee (e.g., a quarterback) wears the shutter glasses and is fully immersed in the virtual play. The trainer (e.g., a coach) sits on the side and uses a laptop to control the training session. A special menu in the Chart Editor program (previously used to model a play in 2-D) provides all control functions for the CAVE. |
From the laptop, any previously modeled play can be loaded in any sequence. The animation can be started, stopped at any point in time, played in slow motion, or played backwards. Each play can be accompanied by sound selected from a library of audio clips. |
Most important are the control functions that relate to navigation. From the laptop, the trainee's viewpoint can be moved to the press box, the sideline, or any other position (on the field or above) that provides a good overview for watching the animation. In addition, the trainee can be transposed to the precise viewpoint of any of the players on either team. This special and most impressive function lets the trainee assume the role of a selected player. In this navigation mode, the trainee's viewpoint is moved automatically with the viewpoint of the selected virtual player while the animation is running. |
During a training session, the trainer can also modify a play very quickly. Since he is using the Chart Editor program to control the CAVE, he can enter the modify mode, change the play--for example, from a passing to a running play--and immediately run the new animation in the CAVE. |
Distributing the play animation over the Web
We have also developed an alternative way to view and explore the three-dimensional play animation outside a CAVE system. Using the VRML (Virtual Reality Modeling Language) standard for the World Wide Web, an entire play can be distributed over the Internet and viewed on the monitor of a laptop or desktop computer. Coaches can provide players with these VRML files and players can study these plays at home or at any other location where a computer is available. |
 The Chart Editor program has a function that automatically creates a VRML file for a previously modeled play. The VRML file contains all the animation features and control functions found in the CAVE application. Viewing the VRML animation, however, is non-immersive, but still useful. As in the CAVE, the VRML user can control the animation and, for example, attach his or her viewpoint to the viewpoint of a selected virtual player. |
The future of virtual reality in sport and science
The pioneering work of the Virtual Football Trainer in the area of virtual avatar animation resulted in new methods for the scripting of complex movements of virtual humans. Based on a two-dimensional path definition, three-dimensional poses, movements, gestures and collisions between humans are created automatically using artificial intelligent algorithms. This innovative concept proved to be of extreme value in our development of this project. |
The modeling and realistic animation of human characters in virtual reality is currently a hot topic. Ongoing research in this area includes combat simulations, training of squad teams for dangerous missions, simulation of accidents that involve people, and the study of human movements in assembly and maintenance tasks. |
For the Virtual Football Trainer, we are currently exploring funding sources for the development of a next generation, commercial product. Possible users include high school football teams, college football players, as well as professional football leagues. Other sports applications that we know are currently in development elsewhere include training simulators for soccer, bobsled racing and car racing. |
For all these applications, including our own, the greatest challenge remains to be the creation of realistic looking movements. Human visual perception can very easily detect movements that are not natural. The most realistic computer animations are based on motion capture data, i.e., data obtained from sensors attached to real humans. This is a costly and time-consuming process. More research in this area is necessary. |
Once a single human has been animated, the next problem is the animation of several humans that interact with each other. This is brand new research territory and, with the Virtual Football Trainer, we have shown a practical solution that includes not only collision detection but also the coordination of individual movements (e.g., a simple handshake between two avatars). The next and ultimate challenge: to enable virtual humans to react to the movements and actions of real humans. |
|
| |
