Virtual Crane demo (VRML2)

The Virtual Crane

Project proposal

(DRAFT)

Author: Silvano de Gennaro <Silvano.de.Gennaro@cern.ch>

Abstract

This document proposes the developement of a web-based general purpose crane simulator within the contest of LHC design. Based upon technology similar to aircraft and vehicle simulators, the Virtual Crane would reproduce realistically the movements, operations and limits of the bridge cranes to be installed in the LHC premises. Using a hardware and/or software reproduction of the future control interfaces, the Virtual Crane operators would be able to interactively drive one or several Virtual Cranes simultaneously, moving objects in the LHC virtual prototypes, and simulating detector assembly phase by phase. These simulations, stored as VRML animations, can later be played back, modified, filmed from different viewpoints, and finally they may eventually be used to digitally control the real cranes during detector assembly.

This document consists of an introduction to simulator technology, followed by a thourough description of the Virtual Crane idea and its implementation phases, and concludes with the its possible future development in the domain of robotic control.

Contents

An introduction to simulators

The Virtual Crane

Robots and Teleoperations

Technology transfer

Conclusions

An introduction to simulators

Machine simulators are today widely used in educational, industrial and military applications, mainly with the purpose of training operators and exploring alternative mechanical strategies at no risk for equipment and humans. With the falling costs of computing equipment and the breakthrough of Virtual Reality applications, these simulators become more and more affordable, and their application is expanding to all fields, with surprisingly high investment return.

Aircraft simulators

Since the first rocket simulators built at NASA during the Space Programme in the mid '60s, aircraft simulators have progressed to a impressive level of realism. Most air companies own these sophisticated and expensive machines, which play a fundamental role in pilot training. Aircraft simulators behave so much like real aircrafts that pilots have a lifelike flight experience. A great part of the pilot training programme is performed in simulators, particularly concerning reaction to emergency situations, which the computer reproduce at no risk. One particularly risky excercise, impossible to train in real life is air combat. Military requirements have pushed research in Virtual Reality further than ever thinkable. This technology is today mature to be transferred to all sorts of industrial sectors.

Vehicle simulators

Once more it was the US Departement Of Defense that opened the road with the first tank simulators. When the first commercial packages became available, the automotive industry followed. Today most car manufacturers include a "virtual car" simulator as part of product conception. Because all car design is performed in CAD, and prototyped in Virtual Reality, the simulator is an easy and natural step to accomplish. Car simulators allow engineers to drive the car before it is built, evaluate its ergonomy, its interior design and even its road behaviour. The simulator cabin is usually mounted on telescopic hydraulic legs that reproduce the feel of suspensions, tyre elasticity and sometimes also acceleration. Sided by high tech realistic graphics, these simulators give designers an excellent immersive experience of their product. Ford and Mercedes use a VR simulator to evaluate instrumentation accessiblity and driver's visibility. Volvo go as far as simulate crash test. But cars are not the only vehicles for which simulators are built. John Deere simulate escavators, Caterpillar have buldozzers and lorry simulators, a Swedish copany, ADtranz, use VR to simulate trains. At CERN, we could enormously improve the LHC detectors assembly strategy by developing a Virtual Crane simulator.


Figure 1: The Caterpillar buldozzer simulator

The Virtual Crane

Environment and needs

Detectors assembly brings enormous space management problems, which are most critical in the case of ATLAS, where the detector will be assembled in its extremely tight pit. In order to avoid space contention and deadlock, each phase of assembly must be carefully planned. CAD models and Virtual Prototypes are already useful to visualize static situations and eventual collisions and intersections. The Robcad software is also being used for path finding within the accelerator modules. However none of these technologies and tools can go as far as simulating appropriately the cranes operations needed to lower and place each part of the detector and its infrastructure.

The VENUS project produced in May '96 some graphical animations which, using the i3D software, demonstrated the possibility of reducing the diameter of the ATLAS main shaft from 26mt to 18mt. This reduction involves pivoting the 200T toroid magnets to descend vertically in the shaft, then pivoting them again in the pit to be placed and assembled horizontally. This pivoting in performed by simultaneous actions by two bridge cranes, both at the surface and in the pit. Available software today cannot reproduce the physics involved in this operation. In order to carefully plan and simulate the complex operations needed for the assembly of each detector part, the VENUS project intends to develop a crane simulator tool.

The crane simulator

If aircraft and vehicle simulators are mainly intended to train a human's piloting skills, or to evaluate ergonomic appreciation of the vehicle before construction, a crane simulator would go even further, becoming a central part of the detector assembly. Designed as a general, configurable toolkit, the simulator could easily impersonate all the kinds of cranes to be used in the LHC complex. Equipped by software and hardware control interfaces, the simulator would give the operator the look and feel of the real crane, operating within the detector's CAD representation. The simulator could offer several degrees of immersion in the virtual environment:



• A web based "on screen" interface directly available through any VRML2 plugin browsers (Cosmoplayer, Live3D, i3D etc.), controlled through the standard mouse/spaceball interface.
• A specialized "immersive" environment based on a set of large screens surrounding the user (cube style technology) and driven through a hardware buttons/joysticks control board.

The basic "on screen" version consists of a functional crane simulation layered over the Virtual Prototypes as available from the VENUS website. It could be developed in 6 months at the most.

The "immersive" cabin will be equipped by a real control board, identical to the wearable radio control desks that will be used in the pits. Coupled with the use of large stereoscopic screens, this is the equivalent of a proper high-end vehicle simulator (see picture above). The crane driver thinks he/she is operating within the real experimental areas, displacing and assembling objects like in reality. In order to enhance this feeling of immersion, the Virtual Crane must react like the real one, considering collisions, weights, elasticity, cable stretch and hook swing. These features can be implemented progressively, according to the "implementation plan" below. The deeper the immersion, the easier and more realistic will be the manipulations and their effects. We have no doubts that such a tool will become immediately useful to the experimental areas assembly planning, even in its earliest phases.

Implementation

The Virtual Crane will be implemented in the VRML2 (Moving Worlds) and Java standards. This means that it will be immediately available over Netscape or Internet Explorer to anybody who wishes to use it. For the later phases we will need to add some C++ modules, particularly for joystic and large screen drivers. This means that the "on screen" version will be largely platform independent (although its usability in a geometry crowded environment will greatly depend upon the host's 3D graphics performance) whereas the "immersive" simulator must be implemented upon a powerful graphic processor (eg. SGI Onyx2 InfiniteReality). Practically speaking, the user will be able to play back the animations over the web satisfactorily on most PowerPC equipped Windows or Mac platform, or modern UNIX workstation, whereas to run the simulator with a likely feel of immersion you will need a very fast interactive response, hence an Onyx.

The Virtual Crane can be implemented in three phases:

Phase 1: "on screen" crane simulation

As said above, this implementation phase is relatively easy and it will allow exploring immediately the first and basic assembly strategies. It will use essentially VRML2 scripts for the functional simulation, and JAVA scripts to represent a button/sliders virtual control board outside the 3D window. This allows free navigation without the risk of "loosing" the control board in the Virtual environment. Navigation will be restrained essentially to the metaphore offered by the local 3D browser. Adaptation of i3D to VRML2 (already started by JF. Balaguer before leaving) will allow the use of a Spaceball and stereoscopic glasses. Apart from the basic crane movements, this phase will feature:
• interactive path recording, playback and editing
• hook collision and grab/release
• floor recognition
• objects stacking by bounding box

Manpower estimate: 6 man/month (tech. student) + 10 days consultancy (JF. Balaguer)

Phase 2: physical behaviour

This phase consist mainly of a dynamics and kinematics study. Objects in the database will be assigned physical properties, such as as weight and elasticity. The crane cable stretch, bounce and pendulum will also have to be computed interactively. Finally we should consider multipoint force and momentum, to cope for natural and controlled rotations and swing. This will push the Virtual Crane into a barely explored field, as most CAD simulation packages (CATIA, Robcad, Megavision) today are limited to simple inverse kinematics. Work in this field should be done by an expert of dynamics and IK simulation. Ideally this should be a one or two year fellowship. Most likely he/she will use and extend an existing collision detection toolkit, then plug the results of calculations into the VRML2 graphic interface. This phase is independent, therefore can be overlapped at any time with phase 3.

Manpower estimate: 18 man/month (fellow or associate) + 18 days consultancy (JF. Balaguer)

Phase 3: immersive cockpit

Probably the best way to implement this effectively is to contract it to a specialized company. Technology in this field is very well advanced, and interfacing any sorts of joysticks, buttons, sliders etc. to a computer is no longer a matter of alchemy. The same goes for large screens, stereoscopic or not. Developing an immersive cockpit may take a specialist a couple of months.

Manpower estimate: contract

Robots and Teleoperations

At this point we have a flexible tool, consisting of cabin that "moves" within the LHC experimental areas, even before we start digging the pits. Sitting in this cabin we can control the bridge cranes through a natural interface, and create assembly strategies for each part of the LHC detectors. These "strategies" are recorded as a path disk files, which can be associated to VENUS' VRML Virtual Prototypes, therefore they can be played back on any computer on the planet. All this is very useful to LHC engineers, for the assembly planning and space management, to the collaboration members, to "watch" their detector being put together, and finally to the crane drivers to train to do their job in a realistic environment. But this is not all. By pushing the project a little bit further, we could at this point capitalize on the work done so far, using these precomputed paths to actually drive the real cranes.

Real time vs delayed teleoperation

Since the analog era, robots could be driven in real time using mechanical arms that could sense the pilot's arm and hand movements and replicated them by the robot. Alternatively, robots could be programmed to repeat a certain movement. The first technique (telecontrol) has been used mostly for interventions in hostile envirments, the second (programming) was usually destined to a production chain. However "often robotics has failed to meet industry expectations because programming or telecontrolling robots is tedious, requires specialists and often does not provide enough real flexibility to be worth the investment. "[1].

Virtual Reality has introduced a totally new way of using robots. By the intermediate of a Virtual Prototype, the operator only works on graphic representations of objects, using "Virtual Tools". The action performed in the Virtual Environment, is then transitted on line (Telepresence) or after validation (Delayed Teleoperation) to a robot, which is programmed automatically as a byproduct of using the Virtual Tool. The result is the elimination (or at least minimization)of the human error factor from the construction budget.

The human factor

This does not mean elimination of the human, but just laying a safety net under him/her. By driving the bridge cranes via paths tested in the simulator, we will reduce considerably the manouvering risks. Instead of handling manually extremely fragile and expensive magnets, often through complex manouvers, which sometime involve more than one crane, the crane drivers could do all the work in the simulator, then let the computer operate the real crane. Human drivers will be there during the real assembly, but they will carry the much lighter responsibility of a START/STOP button.

Technology Transfer

The Virtual Crane is undoubtely an extremely appealing product to a crane manufacturer. The ability to simulate "in vitro" hazardous manouvers, to be later performed in real life by a computer with its digital precision, is a major argument, and could make a real marketing difference. It is quite likely that we will find partners and collaborations within the heavy transport industry. Third party or European funding is also very probable.

Conclusions

The project presented in this paper is consistent and general. Its technology is available, acquired and mature for implementation. Like machines have replaced tedious, heavy or hazardous manual work in industrial environments, simulators are today being used to program these machines effectively and safely. Effectiveness and safety are the two major keys to the successful assembly of the LHC experimental halls. The Virtual Crane implements both.

References

[1] Cannon DJ. & Thomas Geb - Virtual Tools for Supervisory and Collaborative Control of Robots. Presence Vol. 6, No 1, Feb 1997, 1-28.

---

Best viewed with Netscape 4.0
Some icons were mangled using

Comments and suggestions
Last update June 12, 1997.
This document is available from http://www-venus.cern.ch/VENUS/virtual_crane.html