# Trial by Fire: Teleoperated Robot Targets Chernobyl

Jeffrey

Pages: pp. 10-14

The blast that destroyed Unit 4 of the Chernobyl Nuclear Power Plant (CNPP) 12 years ago prompted a firestorm of scientific, technological, political, and economic proposals for managing the worst nuclear accident to date. Following a meeting of the G-7 nations—the United States, Canada, Britain, France, Italy, Germany, and Japan—and Ukrainian representatives, the US Department of Energy (DOE) and National Aeronautics and Space Administration (NASA) organized and funded a "dream team" of experts in robotics as well as computer hardware and software for the "Pioneer Project." Pioneer is a specialized, tethered, bulldozer-like robot equipped with stereo vision for real-time 3D mapping, a core-drilling and sampling apparatus, and an array of radiation and other sensor tools for remotely investigating Unit 4 (see Figure 1). The team has scheduled Pioneer's deployment at Chernobyl for November 1998.

Figure 1   An early 3D model of the Pioneer robot designed to investigate Chernobyl's Unit 4. (Courtesy of RedZone Robotics)

## DISASTER STRIKES CHERNOBYL

People around the world responded to the Chernobyl blast with horror and fascination. We watched telecasts of the burning reactor site, learning of the 32 who died immediately in the explosion and the countless thousands exposed to and injured by escaping radiation. We witnessed the evacuation of the surrounding area and worried over the growing cloud of radioactive gases and dust particles. Thermal plumes as high as 10 kilometers carried radioactive material across portions of Ukraine, Belarus, and Russia. Smaller amounts reached Western Europe, and fallout was detected in the United States.

Ukrainian authorities asked to use robots developed at Carnegie Mellon University (CMU) for the Three Mile Island accident. Cold-war politics blocked the loan. Thereafter we heard that the danger posed by the cloud had passed and the site had been successfully contained under a concrete roof or "sarcophagus." Studies of the surrounding environment and medical histories—for local inhabitants as well as those living downwind—would go on for years, as would the inquiries. But the damage was contained, said Ukrainian officials, and Unit 3 continued to operate and produce substantial revenue. Evidently, Chernobyl no longer merited worldwide public concern.

## MORE TROUBLE BREWING

Not exactly. A serious turbine building fire five years later shut down Chernobyl's Unit 2 reactor. Unit 1 was closed in November of 1996. Now it turns out that the Unit 4 investigative effort has labored under severe limitations, and the containment effort still faces serious problems. Unit 4 remains too hot to handle—for man or machine.

The several-hundred-ton sarcophagus (see Figure 2) was fabricated and installed from a remote distance and with understandable haste, resulting in a faulty seal. Substantial gaps—originally as much as 1,000 square meters, now reduced to about 100 square meters—weaken this roof's effectiveness. Worse, although designed for a 30-year useful life, the roof is deteriorating more rapidly than planned. Its collapse would again release large amounts of radioactive gas and dust. Finally, radioactive water leakage from rain getting through the roof and from internal condensation poses a threat to groundwater.

Figure 2   A photograph of the Chernobyl reactor with the sarcophagus in place on Unit 4.

## THE PIONEER PROJECT

The DOE and NASA funded the project's $2.7 million budget from foreign aid, nuclear safety, and nuclear nonproliferation budgets. The two agencies assembled the most sophisticated talent and resources available: DOE's Lawrence Livermore Laboratory; NASA's Jet Propulsion Laboratory and Ames Research Center; the University of Iowa Graphic Representation of Knowledge (GROK) Lab; Carnegie Mellon University; RedZone Robotics; the Westinghouse Science and Technology Center; the National Robotics Consortium; and Pacific Northwest National Laboratory. Silicon Graphics (SGI), Sense8, and Giant Networking contributed technology to the other players. On 11 May 1998 Silicon Graphics (SGI) hosted a meeting at the Government Simulation and Imaging Conference in Arlington, Virginia. Presenters included Maynard Holiday, principal investigator on the Pioneer Project for the DOE, and Alexander Ivanov, division director of the Interbranch Scientific and Technical Center at Chernobyl. Ivanov described the manned and unmanned exploration and monitoring solutions previously undertaken at Chernobyl, as well as the technological advancements implemented with each generation. He then identified remaining issues for investigation, including those raised because the sarcophagus rests on a structure severely damaged by the blast: • The rate of the roof's deterioration • The amount, composition, and location of melted nuclear fuel or "lava" • The specific topology and condition of the various floors in the Unit 4 building • The amount of radiation absorbed by the walls and roof, and the effect on structural integrity • How long the damaged Unit 4 walls can support the sarcophagus • The remedial options available to shore it up Of more than 190 tons of nuclear fuel originally in the reactor, more than 10 tons have not been accounted for. After the blast, the nuclear reaction melted the fuel into a molten state. As this lava burned through the containing area, it combined with the building matter, which diluted the fuel and stopped the meltdown. The Ukrainians managing the site know the lava puddled in more than one location, but don't know its concentrations ( Figure 3). This raises its own questions: • How much radioactive material remains in the reactor? • What is its composition, concentration, and status? • How much water leaks in or condenses inside, and how hot is it when it drains out? • How much of that water will make its way into groundwater? • Does the composition of the lava vary from place to place? • Given the threat of a future nuclear reaction from this lava, and Unit 4's inability to control or contain such a reaction even if not explosive, what is its probability and where/how would it occur? Figure 3 A side view of Unit 4 following the blast. (Image courtesy of Alexander Sich) The Pioneer scientists believe the answers to these and many other questions—and the resulting remediation strategies—are best accomplished with telepresence using the next generation of telerobotics. This assumption underlies the Pioneer Project. Delay or inaction is not an option. In 1986 the hottest spot in Chernobyl's core emitted 10,000 rads per hour, with current estimates still running one-third to one-half that amount. (The term "rad" stands for "radiation absorbed dose." The US level for acceptable human exposure is 2 rads per year.) Despite the health risk, Ukrainian workers conduct containment and remediation efforts by going directly into contaminated areas. ### The robot The robot design team faces a major problem: the severe radiation inside the reactor would destroy any robot deployed to date. Radioactive particles attack the electrical components, including the vision systems and the onboard computer, and freeze the joints. Of the four types of radiation present at Chernobyl, alpha can be shielded with a piece of paper, beta with a concrete wall, gamma with lead, and neutron not at all. Daryl Rassmussen, a researcher at NASA Ames Research Center in Mountain View, California, referred to a German robot sent into Chernobyl about a year ago—it lasted seven minutes before the radiation shut down the robot's computer and it joined the contaminated debris. Mark Rowland, a physicist at DOE's Lawrence Livermore National Laboratories in Northern California, wrote the proposal that created Pioneer. It differs from earlier attempts in design and materials. Plastics and polymers that provide light-weight strength for normal robots become brittle and fail when bombarded by radioactive particles. Also, as pointed out by James "Oz" Osborn, senior project scientist at CMU, radioactive particle bombardment turns camera lenses brown. The usual solution of encasing components in lead won't work, since it would add 150 pounds of weight to protect a volume the size of a shoe box. The solution is to place many of Pioneer's components in a nearby lead-lined room and connect them to the unit with a tether. Still, this doesn't protect against neutron radiation. RedZone Robotics, a Pittsburgh-based company specializing in robots to explore and clean up hazardous waste, is developing Pioneer. Founder William L. "Red" Whittaker conducted robotics research at CMU and developed the robots used in the Three Mile Island accident. RedZone was founded as a private company to build robots, initially for Chernobyl. When permission to export the TMI units was granted three years after the Chernobyl blast, the issue was moot: the TMI robots were not designed to contend with blast wreckage and radioactive lava pools. Begun in fall 1996 at a reported cost of$700,000, Pioneer is an adaptation of RedZone's Houdini robot used for investigation and cleanup at the Oak Ridge National Laboratory in Tennessee. A small bulldozer four feet long, three feet wide, and three feet high, it weighs about 1,100 pounds and is tethered to a cable about 100 meters in length (see Figure 4). It is designed to turn within its own radius and to navigate stairways and debris. If Pioneer gets stuck, the tether can withstand up to 3,000 pounds of pull to facilitate retrieval. Pioneer differs from Houdini in that it has no hydraulics. CNPP officials, extremely sensitive to fluid spillage and migration, insisted that Pioneer be entirely electrical.

Figure 4   Pioneer looks like a small bulldozer tethered to a cable with a drill in front. (a) Blueprint of the robot (courtesy of RedZone); (b) a 3D rendering of the robot in the simulated reactor environment (courtesy of Alias|Wavefront, an SGI company).

Pioneer carries a 3-foot, 150-pound drill that will bore into the reactor's walls and lava deposits and retrieve samples. Fred Serricchio, an associate engineer in the Guidance and Control Analysis Group at Jet Propulsion Laboratory in Pasadena, California, and Ali Ghavimi, a senior analyst in that group, are building the drilling unit. This unit requires real-time remote control—if the drill hits rebar while drilling through concrete, someone must be able to stop it before the drill bit breaks. They expect samples to show where the concrete was reduced or crumbled from the blast, and where it has become brittle from exposure to radiation. The design for the drill derives from JPL's general research to design a zero-gravity drill for use on space probes to someday drill on comets or asteroids. How the drill's torque affects the parent platform also poses a problem. Pioneer will be the first mobile robot equipped with such a drill.

### The cameras

Mounted on Pioneer's mast are two camera systems, one to provide a video feed to the teleoperator (the "driving camera") and the second to generate 3D meshes and texture maps (the "mapping cameras"). Pioneer's driving camera—a Viticom 2 radiation-hardened camera manufactured by I.S.T. Rees—has an estimated cost of $30,000 and will function for the life of the project. The tube-based design provides color NTSC output at standard 640 × 480 video resolution. Pioneer's mapping camera consists of a three-CCD camera array encased in an L-shaped box about 1 1/4 inches thick. Each camera will have a 8-mm focal length lens which, combined with a 1/3-inch CCD sensor, will result in an approximately 35-degree field of view. The system provides remote user and computer control of the pan/tilt unit on which the cameras are mounted (digital sensors indicate pose of the pan/tilt to within one degree). These standard black and white video cameras—currently made by Hitachi for about$400 each—have the auto-iris and auto-gain features disabled for image consistency. Although lead-shielded, they are not radiation-hardened and will fail from exposure to the intense radiation, requiring replacement.

The team prefers black and white over color for the triplet stereo image generation because it gives the most accurate composite image, avoiding the registration problems of reassembling three color channels on each camera and compositing them. A framegrabber on the PC will acquire images at normal video resolution, shipped via sockets to the SGI for stereo processing and display. Each stereo image triplet will be evaluated for suitability (for example, density of the resulting depth map) by the stereo code. Images may need to be retaken with user adjustment of the lights if the depth map is too sparse or if the mesh merging step fails to find an adequate registration of adjacent range maps.

Accurate sensing requires precise calibration of the stereo cameras. Geometric calibration consists of correctly modeling each camera's optical path, helping operators correctly predict the 3D locations of objects from their 2D images. The best way to accomplish this calibration is to take a picture of a 3D target with a known shape so that feature points on that target can not only be easily found in a 2D image, but can also be easily mapped to their known 3D locations. JPL is working with Spacecraft Specialist to provide a large calibration target (three faces, each face one square meter in area) for this purpose. Scientists in Ukraine will use this target to recalibrate the stereo cameras upon delivery and optionally during deployment within the Chernobyl facility should the cameras become misaligned.

While black and white imagery is preferred for generating 3D meshes, the real-time texture mapping must be in color. CMU's Osborn explained that this aspect of the design remains under evaluation for now. The current approach applies separate red, green, and blue filters to the lenses on the mapping cameras. Compositing these results should avoid registration problems. But the team is evaluating the alternative of using a separate color camera.

A second camera issue requiring a solution involves how to adjust the brightness/contrast in the black and white images. The grayscale values in the image affect how Pioneer constructs and maps the mesh. Pioneer has an array of dimmer lights to compensate for the bad illumination; only the remote operator can adjust the camera brightness and contrast. But the design team recognizes this may be too much to require from even the most conscientious operator and is looking at ways to automate this function.

### Using the robot

RedZone's engineers designed Pioneer with a long tether for operation at some distance from the reactor room, based on their own inspection of the site. In fact, the Chernobyl operators intend to reclaim, clean, and lead-line a room two or three doors down the hall from the reactor. This room will serve as their base of operations. Measured by US nuclear practices, this proximity presents a substantial health risk—a sacrifice that has over time become accepted by Chernobyl workers.

In practice, Ukrainian technicians will physically carry the robot into an outer chamber, quickly set it on the floor and retreat to the lead-lined room. The robot can then proceed to the most dangerous part of the plant, near the reactor. Asked how long Pioneer will last, Bruce Thompson, vice president of engineering at RedZone, said it all depends on usage: the design specification is total exposure of 1 million rads. Given the conditions, Pioneer should function for at least year. RedZone will deliver one unit; there is no backup.

Thompson explained that the robot is built to inspect and sample the 2,500 square meters of Chernobyl. It will then be retrieved, torn down, and serviced near the cleaned, lead-lined operator's station on site and redeployed in the reactor room. The designers anticipate certain unshielded electrical components will fail under the intense radiation and be replaced at an acceptable cost. Because evidence suggests most radiation within the sarcophagus is particulate, CNPP operators believe a great deal of the radioactive contamination can be cleaned off the unit when it's serviced.

### Real-time 3D visualization

NASA, working with collaborating agencies, universities, and corporations, adapted and enhanced the computer-based 3D visualization technology originally developed for last year's Mars Pathfinder mission. The robot has an array of radiation-hardened sensor equipment (for radiation levels, temperature, humidity, and so forth) developed for it by the Westinghouse Science and Technology Center at Churchill, Pennsylvania, which has experience with radiation.

Software applications developed at NASA Ames, CMU, and the GROK Lab using WorldToolKit, Sense8's C or C++ 3D virtual reality software development environment, carry out Pioneer's mapping function. Bringing VR imaging to the Chernobyl robotics effort offers two capabilities. First, as 2D stereo image data comes in, the remote operator can respond with course corrections or other actions. But you could do this with a video feed, right? Why would you need real-time 3D using three Octanes and an Onyx2 supercomputer (all donated by SGI)? The second capability provides the answer: while viewing the video feed, Chernobyl's operators can use virtual cameras to examine any area from any angle or any distance, or from multiple, simultaneous viewpoints. Multiple perspectives or real-time flythroughs are possible only using VR techniques. Most important, a reality-based virtual model gives immediate access to all the data at once, regardless of the robot's position or status. As explained by Linda Jacobson, SGI's VR evangelist, VR technology applies where mission-critical data must be analyzed in real time. Chernobyl definitely fits this description.

One Octane will sit in the lead-lined remote operating room. The operator will use it to track the robot's location, convert 2D stereo video images to a 3D surface mesh, map video images onto the surface mesh, display the mapped mesh within the context of the rest of the building, and analyze core samples for structural stability. A second Octane, housed in Chernobyl's administrative building, will be networked with the first to provide backup, service, and support. It will also facilitate inspection and refinement of the models, provide a platform to share data with government officials, scientists, and others, and serve as a networking hub connecting Ukraine to the US. The third Octane will remain at the University of Iowa, connected to Ukraine by satellite link as a gateway for other Pioneer partnering entities. It will provide a platform for code development and enhancement, serve as a copy of the computer deployed on site, and display the models generated by Pioneer. The Onyx2, also located at the University, will facilitate real-time interaction with the photorealistic models to generate flythroughs of the site to assess structural integrity and other factors.

## POLITICAL CONSIDERATIONS

In December 1995 Ukraine signed a memorandum of understanding with the G-7 nations to close the Chernobyl plant by the year 2000. As part of that memorandum, the G-7 nations agreed to help remediate current risks at Chernobyl, support energy efficiency, and help alleviate the socioeconomic impacts of Chernobyl's closure. Representatives from Ukraine and the G-7 countries met at Denver, Colorado in June 1997 where the G-7 pledged to fund the $750 million estimated cost of closing the plant. About$300 million has been raised. The Pioneer project, although not proposed as part of this plan, was inspired by it.

Right now, the cost of maintaining the tomb drains Ukrainian cash resources. Reportedly, when told the Pioneer cost would come out of the US's $78 million portion of the G-7 contribution, the Ukraine asked Washington to cancel Pioneer. In response, Washington agreed to fund Pioneer on its own, separate from the shutdown project. Reports in the Western media give the impression everyone agrees Chernobyl should be closed. However, Unit 3's power production and revenue stream have produced a constituency cool toward the US robotics initiative and proposed shutdown. The official Chernobyl site ( www.chornobyl.org) contains the following statement: The problem of closing [the] Chornobyl NPP is very difficult and complicated. There is an opinion that there are not any technical and economical reasons [for] closing it and that it is [a] political problem. Since the 1986 accident, the safety of CNPP has been considerably upgraded and the root causes of the accident have been remedied. And now, after these renovations, specialists believe that repetition of the 1986 accident is impossible. Also, CNPP is a well-organized business which earns annually$300 million of income. But in spite of all [this,] countries of the G-7 are trying to close CNPP, because if they manage to do it, they will [reduce the] fear of [their] citizens and therefore they will [be able] to build [new] NPPs in their countries. But countries of the G-7 didn't carry out all [their] obligations [made to] Ukraine, so it is possible that the Unit 2 of CNPP, which was closed in 1996 at [the] insistence of countries of the G-7, will begin its work again after years of interruption.

Obviously, politics and economics will play their part in what happens at Chernobyl in the future. Everyone involved in the Pioneer Project hopes data from the robot will indicate whether the sarcophagus must be repaired or rebuilt, how long the underlying structure can sustain its weight, if the lava poses a danger of critical nuclear reaction, and specifically how water leakage and condensation threaten ground water. Presumably this data will shape strategies for avoiding a roof collapse and preventing future nuclear reactions and ground water contamination. But given the economic and political realities, only time will tell how important a role Pioneer will play in the course of remediation and Chernobyl's ultimate fate.

## URLs Related to the Pioneer Project

Official Pioneer Project Page (Carnegie Mellon University)

GROK Lab Pioneer Page

NASA Intelligent Mechanisms Group

RedZone Robotics, Inc.

Silicon Graphics, Inc.

Sense8 Corporation

JPL Machine Vision and Tracking Sensors Group

Department of Energy