Consultation
As our report is supposed to be an overview of the problems and issues related to this domain, rather than an in depth specification, our use of this method of research was minimal. However, when we required information about the Camcopter[4] discussed in the design specification, we did require additional specific information from an expert who was partaking in the development of the system. In general however, we found that the information extracted from the web was more than adequate.
Description of Process
Group organization
As our project group consisted of six people, we collectively gathered a wide variety of information on the landmine problem. We started out by individually gathering as much information on sections of the problem that we had designated to each other. We then had regular group meetings to discuss the information we had found. We then evaluated this information to allow us to develop a better understanding of the problem, and also to enable us to determine which data was most relevant to the problem domain.
Once we had a clear definition of the problem, we were able to split ourselves into three groups with each group of two people working on a particular area of the problem, as well doing more research on that particular area.
In more recent group meetings we brought all our information together and discussed any problems that we had encountered. This allowed us to formulate ideas as a group and to help each other with problems that we were encountering. We found this very helpful, as it meant that all members of the group were kept up to date with the entire problem, rather than just their own area. After splitting up again to work on separate areas, we again reformed into one group to help formulate the optimum way to report our findings.
Development of Analysis
Throughout the project we frequently encountered minor problems that we could not solve, so we had to find an alternative way to work around them. When we initially started the landmine problem we approached it as one problem area, the removal of landmines from a particular area. We soon realised that problem consisted of two comprehensive areas, detection and detonation. From this point on we focussed our research around these.
Initially at the start of our research we began to concentrate on the expert system and heuristics side of the problem, or concisely, the ‘AI’ aspect .
We developed two complex algorithms that attempted to optimise the time our proposed robot fleet took to search and destroy. We also used a variation of the ‘travelling salesperson’ algorithm to get each robot into an optimised position for detonation, after the detection phase had been carried out.
After further thought and advice we decided to take a different route, and instead of concentrating on the heuristic and mathematical side of the problem, we decided to put more effort into the research of existing and futuristic landmine search and detonation methods that we could use. As the report specification allowed a maximum of 20 pages, we only give a summary of the algorithms that we had previously intended to develop fully.
Specification and Design
Defining the search area
The first factor that we wish to take into account is the size of the specified search area. If the search area consists of hilly, vegetated, rugged terrain, we would be looking at a small size, whereas if the area was flat, desert-like terrain, we could probably handle something larger. Obviously, the number of robots at the system’s disposal will greatly effect this decision. We must also consider the fuel/power source of each robot. The larger the power source, the further it can travel, which would suggest a larger search area. Also, the rate at which the detection mechanism can search is of importance. All these factors must be taken into account. At an estimate, our search area would be between 200 x 200m, and 1km x 1km, depending on the previously mentioned factors.
An important requirement is the identification of priority areas. Despite the automation of many of the de-mining tasks, it must be accepted that there will be at least some element of human interaction. One such instance might be when priority areas have to be determined. Automating this process may be a tricky task, whereas a human can easily identify areas that might require special attention. If the human did this task, it is likely that location values representing these priority areas would be passed to the system before robots have even entered the search area. If the robots/centralized manager determined the priority areas, it is obvious that they would have to be done after an initial mine detection phase had been carried out. Factors that might determine priority areas might include the distance an area is from humans or to a lesser extent, rare species of wildlife or animals. If humans traverse frequently across a particular area, this should be considered. If any historical knowledge of a search area where military conflict occurred is known, this information could possibly be used to determine where a high concentration of landmines were most likely to have been planted. An example would be if a particular army in battle were retreating over a bridge, landmines are likely to have been planted in the vicinity of the bridge to hinder the attacking army’s movement. If a historian had some proof that something like this could have actually occurred within our search domain, this evidence could be taken into account, when priority areas were being determined. Obviously if the determination of priority areas was automated, the previously mentioned factor would not apply.
In terms of mapping our domain, a ‘square’ or regular search area is also preferable, because it is generally much easier for an automated robot fleet to traverse a regular shaped area rather than one that would have an ‘irregular’ appearance on a map. A regular-shaped search area also endorses thoroughness, as if the search area was irregularly shaped, there might be indecision amongst the robots about what areas have and have not been searched, although one might argue that this is always a problem. Also, it would make the representation of locations and distances that bit easier to apply to the system, although it is very possible that some experts may disagree with this assumption.
How locations and distances relating to our domain are represented on our system is an important issue. Principles used in navigation circles also apply here, except accuracy is of more importance in this domain. A system invented in the 1950’s, known as electromagnetic distance measurement (EDM),[11] is based on a principle known as triangulation, and provides the basis for today’s Global Positioning Systems (GPS)[13]. An earth-based GPS receiver can determine its geographic location from information it receives from several dedicated GPS satellites in earth orbit. There are presently 24 NAVSTAR satellites that broadcast GPS signals globally. The use of regular GPS will not provide us with the required accuracy for this domain, as it is only accurate to about 10 – 40 metres. A system known as Differential GPS improves this to around 1-3 metres, however this still isn’t sufficient [8]. However, the emerging development of Carrier Phase GPS, allows us to obtain sub-centimetre accuracy [13], which would be required when locating and clearing landmines. As expected, this is a very expensive technology, but like anything else, as the demand for it emerges, the cost should drop considerably. GPS can represent specific locations using (x,y,z) coordinates, where x and y determine its geographic position, and z determines its altitude relative to average sea level values. These coordinate values can then be represented to our robots/centralized system.
Knowledge Bases
Expert systems must be knowledge-rich even if their methods are poor.1 A good knowledge base is the key to a good expert system. In the landmine problem we would require many varying knowledge bases. Examples of these might include terrain types, mine types and weather conditions. When detecting or destroying a mine, the system will need to use its array of knowledge in order to make various important decisions.
Being made immobile by either falling a great distance, sinking or getting stuck in unstable terrain is a major problem for a search and detection system. In any search space it is likely that the terrain will be of a diverse nature.
The topology of all countries varies enormously and because landmines have been used in up to 69 countries,7 this means the knowledge base of terrains will differ for each country and climate. Terrain is also one of the main factors in the planting of landmines. Terrain can be rocky, rolling or flat, and landmines cannot be planted in concrete, rock or deep water. It is essential that the search and detection devices will be able to distinguish between different types of terrain.
Another consideration would be water. Landmines have previously been planted in water that is shallow enough to wade through. There are up to fifteen various water mines in existence today. An example of such a landmine would be the ALCM-82 shallow water mine, used in North Korea2. A detection device must be able to distinguish between water that is shallow enough to wade through and deeper water.
Another consideration for the knowledge base of detection systems would be forest and areas of high vegetation. Any system must know how to navigate these areas successfully. Vegetation in some areas may have to be cleared before any detection can be carried out. The knowledge base must be able to distinguish between areas of not so dense vegetation that can easily be searched, and vegetation that must be cleared. Then the system, with the help of its knowledge base should be able to determine the best way to clear it. A machine produced by Mine Tech, known as the Agribush Flail, 3 can be used to clear vegetation in mine fields. A knowledge base of soil would also be needed, and again it must be able to distinguish between these types.
Probably one of the most important aspects of the knowledge base of any detection and detonation system would be its database of mine types. Mine types come in three main types - antipersonnel landmines, anti-tank landmines and blast antipersonnel mines. Anti-personnel mines with fragmentation are only partially dug inside the ground. In some cases, they bounce up before exploding, or they might explode in a certain direction. Fragmentation mines are frequently triggered by wire and are lethal within a radius of about 30m. Anti-tank mines are packed with explosives to disable any tanks unfortunate enough to come across them. They react on ground pressures of 150-300 kg or by induction. Blast antipersonnel mines include less than 100 grammes of explosive. As they are small and easy to manufacture, they are currently the most plentiful type of mine. When surrounded by grass, they can be difficult to see. Blast mines can be detonated by a ground pressure of about 10kg/dm2, usually using trip wire. These mines are not designed to kill, but to badly maim4.
A detection system must be able to tell the difference between these three main types of mine. Once it has identified the main type of landmine then there are up to 700 different mine types within each category. This adds up to over 2000 different types of landmines with each mine producing country making up to 100 different mines each. The former Soviet Union has produced up to 115 different mines alone5.
All these land mines vary in case material, shape, length, height, diameter and the material they are made of. Each mine has also has a different effect such as blast, fragmentation, side attack, direct fragmentation, shaped and dual shaped charge.
A good knowledge base of landmine types is especially important for the detonation of landmines. The detonation of an antipersonnel mine that can spread up to fifteen meters of shrapnel can have a detrimental effect for the surrounding environment, making the land useless until all the remaining pieces of metal have been cleared. A good system must be able to identify each landmine type and associate an appropriate way of detonation with each mine type. The options here include moving the mine to a safe place and detonating it there, or to detonate it on the spot.
One major problem posed in landmine detonation and clearance is the high false alarm rate which most of today’s systems encounter. The US army’s false alarm rate is one false alarm per 1.25 square meters6. This is a huge problem, as any humanitarian de-mining technique’s main goal is to be as accurate as possible. When accuracy is taken into account, every piece of material that could be a landmine must be treated as such.
One way to reduce the false alarm rate of a system would be to include a good knowledge base of materials that could possibly be landmines. This knowledge base would have to contain all known materials that any existing landmine consists of. These include metal, plastic, wood, concrete, glass and ceramic. However, it would be important that the system would be able to determine whether a detected piece of metal was actually a mine or not. One thing that all mines have in common is that they are all packed with TNT or other types of explosive. If an explosive material is detected then a detection system can almost guarantee that it has discovered a landmine. If no explosive is detected, then the piece of metal may simply be a piece of shrapnel from a mine that was previously detonated. Of course, the piece of metal could also be there for other reasons.
Design of Robots
In this section we must consider possible design principles for our robot fleet. These include issues such as:
- how the robots will be organized in order to work as a unit
- how the robots will communicate with each other
- detection and destruction mechanisms
- protection of robots from unintentional mine blasts
Robot Configuration
Obviously, how the robots communicate with each other is very important. We could adopt a policy whereby the robots communicate directly with each other, constantly notifying each other of developments in their particular area of the search space. Alternatively, we could also have a centralized system, where there would be one automated robot manager in the vicinity of the search area, which would designate tasks for each robot, while monitoring the status of all robots working in the search domain [1]. With this system, the robots may not need to communicate with each other, and they would only do what they are told to do by the central manager. So for example, if two of the robots were in danger of colliding with other, the centralized manager would send a warning to each of the robots, rather than the two robots warning each other.
Aerial search and destroy technologies
If we were to adopt an aerial ‘search and destroy’ strategy, we would have a couple of options. After our research on balloon technology, we realized that this is currently not a feasible solution, although technological advances are being made in this area. Virgin’s multi-millionaire owner Richard Branson is currently campaigning for corporate support to finance research in this area.[2] Branson believes that balloon technology, used in conjunction with the latest radar technology that is capable of detecting and mapping mines at 100m/s, could possibly result in a major breakthrough in fighting the landmine problem.
Currently, the main problem with balloon technology is its inability to cope with windy conditions, although Floatograph Technologies in Napa, California are making major strides in this area. They claim that they are currently developing a balloon that can handle hurricane force winds [3]. It wouldn’t be surprising if more significant breakthroughs were made in this area over the next few years.
However, there presently is a better solution – known as the ‘Camcopter’[4]. This is a remote controlled helicopter that has been specially designed for identification and detection of landmines. It contains an on-board GCD camera and infrared sensor, which is used in conjunction with a Ground Control System. During testing, the ‘Camcopter’ performed at impressive cruising speeds of 90km/h, and reached altitudes of up to 1700 metres, albeit in favourable weather conditions. It has a maximum fuel capacity of 5 to 30 litres, and an average burn-off rate of 4 litres per hour. Like most anti-mine systems being developed, it performs best in flat, uninhabited areas where vegetation is minimal, although it is capable of working well in less favourable conditions.
Although balloon technology could possibly be a more cost-effective solution, the ‘Camcopter’ appears to be the best aerial mine location system present at the moment.
Physical Communication of System
How each of the robots will physically communicate with each other, or with a centralized coordinator is another issue. Infrared is currently being used to allow robots to communicate and identify each other’s position, but this is not applicable here, because it’s range wouldn’t be sufficient for this domain. A system based on broadband cables or fibre-optics is immediately dismissed because of the simple reason that we can’t allow the presence of long cables all over our domain, risking the detonation of landmines, which could possibly result in the destruction of one or more robots. So therefore, we need a wireless solution. In tandem with this wireless solution, we could use a socket-based technology for robots, known as RoboComm. It is basically aimed at simplifying asynchronous computer-robot/robot-robot communication, using standard socket principles in conjunction with Java based objects [1][5]. Using this system, robots can broadcast to the centralized server, or to each other, depending on how the system is configured.
To compliment this we need some sort of wireless medium, probably radio of satellite. Examples of radio-based modems would include the YDI series, most notably the YDI Model 192MM[6], and the Paradise Datacomm P400-480 series provide more than adequate satellite coverage[7]. A radio modem installed in each robot would ensure adequate coverage, but the much superior data transfer rate of the satellite modem may be required, depending on the complexity of the design. A brief example would be if human analysts located a long distance away from the search wished to view the robots in the search area, a camera could be installed on some or all of the robots. Using our satellite link, we could transmit the pictures to probably anywhere in the world. Radio waves would not provide us with this facility. However, radio technology is currently much more accessible than satellite in the world’s poorer areas, where coincidentally the vast majority of the world’s landmines are situated.
Neutralization and destruction of landmines
During our research we have discovered three main methods to deal with this area of our study. The first, ‘Mine Marking and Neutralization Foam’, provides a method to safely mark and remove exposed landmines if neutralization is not possible of preferable at a particular point. Polyurethane foam is applied to the landmines. With its bright colour easily marking the mine, the foam then hardens, impregnates the exposed parts of the mine, rendering the fuse inoperable. This hardened foam then prevents detonation of the mine even if it is accidentally landed on, or stepped on. The foam acts as an adhesive to stick a rope to the mine so that it can be safely removed to be later detonated in a controlled explosion. Although this method may be successful in humanitarian efforts involving human de-miners, it is obviously unsuitable for our requirements as it only deals with mines that are exposed on the surface of the earth.
The second method is a device known as an ‘Explosive De-mining Device’ (EDD). This device is a tripod mounted shaped charge integrated into a fixed-time delay fuse assembly. After rigourous international testing, these devices proved themselves to be very effective, mainly because their ability to destroy plastic, metal and wooden mines at a ground depth of up to 22 inches. Advantages of the device include the fact that it is simple to operate, and is sufficiently stable to be moved and stored as a class 1.1D explosive [1]. The main disadvantage of these devices is that even though they are simple to use, the operation of the device is still hazardous as it involves the use of shaped charge explosives. Currently there is no way of electronically detonating the device, and the time delay fuse could prove dangerous should our robot get into difficulty. Detonating metallic mines also creates more metallic fragments to further confuse later de-mining operations, and may also hinder the use of such land for agriculture in the future.
A third option, which is in our opinion the most suitable one, is a de-mining system that consists of a Chemical Neutralization Device. This involves the remote control firing of a bullet through a “gun” into the mine. These bullets contain a capsule of chemicals such as diethyl amine, diethylenetriamine and diethylzinc. These bullets penetrate the mine casing and initiate a reaction within the mine that will burn the explosive charge. They work even if the mines are buried and air is not required to sustain the combustion within the mine as the explosive itself has oxidizing power in NO2 groups[2]. This then disables the mine allowing it to be extracted form the earth and moved to a safe location for a controlled explosion to ensure that the explosives are completely one hundred percent destroyed. The advantage of these chemicals is that they don’t detonate the explosive. We would recommend the use of such devices because there are no harmful environmental side effects associated with this system. The amount of chemicals used is very small and they are consumed during the neutralization process [3]. The benefit of a system that destroys mines without leaving fragments behind is hugely significant for us.
However, problems do exist, because at the moment the chemicals must be tailor mixed depending on mine case thickness. The de-miner must therefore know the exact type of mine that they are dealing with, and what mix formula to use on it. The special handling and storage requirements of the chemicals further increase the complexity of the system.
The aim for future developments [5], is to have one chemical and one delivery system to neutralize all mines, elimination of the need for specialized storage and handling requirements, and to make the system simpler, less expensive and disposable.
Detection methods for Landmines
Infrared Detection
As mines release heat at a different rate than their surroundings, we can measure the thermal contrast between the soil close to a mine and the rest of the soil using Infrared detection. This can be done during natural temperature variations in the environment. The main problem with this however is that infrared can depend heavily on the environment, and is known to work poorly in hot conditions. The presence of thick foliage can also cause problems.
Image from Infrared during the day Image from Infrared during night
Ground penetrating Radar (GPR)
This is also a very expensive method of mine detection. This system works by emitting an electromagnetic wave into the ground, through a wide-band antenna. Reflections in the soil caused by dielectric variations such as the presence of an object are measured. By moving the antenna, it is possible to reconstruct an image representing a vertical slice of the soil. A problem here is that resolutions needed to cope with very small objects enforces the use of frequencies of some GHz, limiting the penetration depth and increasing the image clutter.
Seismic Wave Analyser
This works by simultaneously using sound waves to create tiny soil disturbances, and precision radar to measure the resulting movement. It uses a transducer to create seismic waves that travel through the soil. The tiny movements in the soil can be detected by electromagnetic waves from a small radar system attached to an airborne robot. This system can also differentiate mines from other buried objects. However, as it is still only in its testing phase, it can only detect ten different types of mine, but with time its developers promise that this will improve.
Microwave Technology
This uses microwaves to detect objects in the ground. The use of wavelengths, allow the system to retrieve an overall size of an object, making it easier to recognise mines. This method can detect both metallic and non-metallic mines. The problems here are that there are many false alarms, and it is also a lot more difficult to detect non-metallic mines. For our system we would recommend the Seismic Wave Analyser. This device presents us with the most positive aspects, with its ability to discard false alarms and categorise mines, but from a negative aspect the range of mines which the device recognises is still minimal. However with time, this technology is sure to develop to the extent that it could be used in an automated system.
Search and Destroy Methods (see diagrams in Appendices)
The use of a system of autonomous robots for de-mining is a more practical solution that that of exhaustive human searching for a variety of reasons, the most obvious being that of safety. An expert system is also able to optimise its search strategies to improve the speed of the operation, and with today’s technology the thoroughness
Although expensive, the loss of a robot is far more desirable to a lost life or limb, and the fact that we are using airborne search robots also minimizes the risk of accidental detonation. Another precaution built into the system will be the assignment of priority areas. These areas will be searched and de-mined first for the safety reasons.
Another big factor when searching will be the speed at which the robots can search a given area, and how efficient the strategies employed are. An example search strategy would be to:
- Divide the priority areas into equal search spaces between the robots.
- When a robot finds a suspect item it will send the coordinates and the depth of the item to the centralised manager where it is recorded.
- Once a robot has finished its given search space it will be instructed by the centralised manager to move to the end of a slower or destroyed robot’s search space where it will start search back towards the other robot.
- Once all the priority areas have been searched and all suspect items are recorded the disposal of these items takes place.
- This algorithm is then repeated for the rest if the search area until all suspect items are recorded from the given search space.
This algorithm can be optimized by dividing the search areas in an efficient manner and also by minimizing the amount of unproductive movement performed by the robots.
Although optimization and efficiency are major issues, we cannot forget the reason for searching is to find mines and to be as thorough as possible when doing so. First of all we can see that the entire search area will be searched if there is least one functioning robot left, and also with modern technologies it is becoming easier to be more accurate when finding mines.
Destruction Method Used:
After the suspect mines have been charted for a given area, the system must be set up for the disposal of the mines. The major issues involved once again are safety, thoroughness and efficiency. The main reason for using robots for the destruction of the mines is to save human life, sometimes at the expense of machinery.
Our robot will have a number of different disposal methods, one of these being that of chemical neutralization. Once the mines are thought to have been chemically deactivated (some of the explosive chemicals may remain), they can be transported to specially designed blast pits on all four sides of the search area by a robot called a ‘mini-flail’ and can undergo a controlled explosion for further reassurance. The blast pits are designed to minimize the damage done to the surrounding search area and to contain any shards of shrapnel. Also for safety reasons the priority area will be demined first.
The strategies the robots use will define their movement, so it is an integral part of the destruction process. Here is an example strategy:
- Starting with the priority areas, we create a shortest path between all the suspect items (similar to the Traveling Salesperson Problem), and divide the path equally between our robots. All paths to and from the priority areas are carefully planned by the robot manager to avoid all suspect items in the normal search area.
- Each robot will attempt to neutralize all suspect items in its path.
- As with the search algorithm, any finished robots will be instructed by the robot manager to help robots that are still working on their path.
- Once all items have been neutralized, all the mines will be individually transported to blast pits, which are on all four sides of the search area, and once there an attempt will be made to detonate them using a controlled explosion.
- Once the priority areas have been cleared they rechecked for reassurance.
- The strategy is repeated for the rest of the search area until all the mines have been cleared.
The use of the TSP to hopefully give us the optimum route that the robots should take so that very little time will be wasted by the robots by doubling back on themselves, or the robots cross each others paths very often.
The thoroughness of the strategy is assured by the facts that the entire search route is rechecked after the mines have been cleared, and also by not only neutralizing the mines, but also by detonating them as well.
Discussion of Design and Conclusion
Here, we attempt to evaluate our system design in an unbiased yet positive manner. The group saw the system’s method of navigating the search area as an important area of evaluation. Firstly, the proposal of an airborne system was a major strength of our design on a number of fronts. Airborne traversal of the search domain is much quicker and safer than land based methods. It is also a much safer method of searching the domain the fact that humans are kept away from the mine-populated areas, and also it reduces the possibility that our expensive robot fleet will be destroyed by accidental mine blasts. Unlike land-based systems, airborne methods do not have to waste time trying to navigate difficult terrains such as areas of forest and extreme vegetation.
The area of detection was recognized as another area where the group made positive developments in, mainly because of the use of the Seismic Wave analyser, which results in extremely low rates of false alarms. This is a major problem with existing systems.
The disposal of the detected landmines was considered to be an area of immense importance. We paid particular attention to the fact that certain disposal methods will remove the mine, but at the expense of the surrounding environment. This is why we have recommended the use of chemical neutralisation to safely remove the mines, as it does less damage to the environment than the other previously mentioned methods. Also, we have taken into account that the mines should be moved to a safe area before being destroyed, therefore reducing the amount of harmful shrapnel in the search area. This helps reduce the rate of false alarms, as well as increasing the possibility that the land may be reclaimed for agricultural use in the future. Also, our proposal for an extensive knowledge base will differentiate between which areas mines can and cannot be planted and will consequently act accordingly.
After looking at all the positive aspects of the system, we must now take a look at the negative side. The robot itself will be difficult to transport, because it is very large in size, and also it is of an airborne nature. This will add to a budget that is already very high. The cost of all the equipment needed to have just one fully functional mine finding robot is astronomical. In this system we propose the use of ten such robots. The inclusion of radar devices, chemical neutralization systems and GPS systems will all add massive expense to the robot.
The chemicals, which are used to neutralize the mines, are highly reactive and require special handling when being transported. Although being an automated system, a certain amount of human intervention is needed, particularly in the assignment of priority areas. The huge overhead in designing a system, which would recognize all these areas, wouldn’t be feasible.
Our ‘camcopter’ system, being airborne, would only be hindered by the presence of trees or perhaps some heavily vegetated areas. These areas would possibly have to be cleared in a safe method, before the system could be used. This is a problem that all de-mining techniques encounter, and is one that would be very difficult to solve in practice.
Returning to the subject of human intervention, we should also state that the system would require a group of expert users highly trained in the functionality of the system and a deep understanding of its background. This would also add expense. It is not realistic to presume that an automated system could handle the entire de-mining process without the presence of humans. It is doubtful that this scenario will ever change.
In conclusion, it is easy to see that cost is a huge problem when designing a system of this nature. These problems prove that it is very difficult to design a flawless system.
Appendices
Landmine Hotspots
Different colors represent the landmine problems in the affected countries of the world
Areas worst hit by landmines
Indicates the areas worst hit by landmines, concentrating on Mozambique
Camcopter
References
Dawson-Howe, K M and Williams, T G 1997 (8-11 July). Autonomous Probing Robots for the detection of abandoned landmines. Pages 51-57 of: Proceedings of the 5th Symposium on Intelligent Robotics Systems (SIRS-97).
http://kzoo.edu/~k98cm01/edwardfeigenbaum.html
[1] Feigenbaum, Edward A. The Fifth Generation; Artifical Intelligence and Japan's
Computer Challenge to the World. Reading, Massachusettes: Addison-Wesley,
1983.
[2]http://www.demining.brtrc.com/MineSearch.ASP
[3]http://www.demining.brtrc.com/R_D/clr_proddet.asp?productid=151
[4]http://diwww.epfl.ch/w3lami/detec/monterey951.html
[5]http://www.demining.brtrc.com/MineSearch.ASP
[6] Science & Technology Review November 1997
[7]http://www.care.org/info_center/sr_landmine.html
1] “Team aware Robotic Demining Agents for Military Simulation” –
Gita Sukthankar +Katia Sycara, Carnegie Mellon University, Pittsburgh, PA
[2] “Virgin’s mine spy in the sky”
[3] Floatagraph Technologies – “Taking Landmine Field Response to New Heights”
[4] “Camcopter – Human Demining Developmental Technologies 1998”
[5] “TeamBots Application : RoboComm”
[6] “YDI Spread Spectrum Radio Modem Range”
[7] “Paradise satellite Modems”
[8] “GPS System Accuracy after 1 May 2000, Some UK Tests”
Ian Strachan
[9] “TODS component listing”
[10] “MK40: The World’s Most Accurate GPS Chart Navigator”
[11] “Modern Surveying”
Encyclopaedia Britannica
=51756#51756.toc
[12] “David L. Wilson’s GPS Accuracy Web Page”
[13] “A GPS Primer”
- published in Cartouche, No. 29, Spring, 1998
Mike’s references
[1] http://pubs.acs.org/hotartcl/cenear/970310/land.html
[2] United States Government, Department of Defence, Scientist Patel, http://www.defenselink.mil/
[3] http://www.demining.brtrc.com/
[1]Kelly, Geraldine – “Anti-Civilian Landmines in Modern Conflict”, University of Limerick, 1999
[2]Sheriff, Andrew Mark – Landmines: A Perspective on a Global Problem, University of Limerick, 1994
[3]Craib, J.Alastair – “Mine detection – the military necessity to render anti-personnel mines non detectable in ICRC. Symposium on Anti-Personnel Mines Report – Montreux, 21-23 April 1993 – ICRC Publications, Geneva.
[4]Africa Rights + Mines Advisory Group (1993) – “Violent Deeds live on: landmines in Somalia and Somaliland” – Africa Rights, London.
[5] de la Billiere, Peter. (1993) “Storm Command: A Personal Account of the Gulf War” – Harper-Collins, London.
