|
|
LANDMINES |
|
 Robots are not a solution to the global landmine problem!I have argued this in two principal papers:
A summary of my arguments follows: Several large research efforts have failed, so far, to develop an effective mine clearance alternative to the existing manual technique. Robots have been tried at great expense, but without success. This paper argues that robots are not an appropriate solution for mine clearance. First, there is little likelihood of sensing improvements in the short term. Second, the huge variety of mines and minefields defies any automated solution. Third, robotic solutions are likely to be too expensive to be practical for humanitarian demining operations in countries like Angola, Afghanistan and Cambodia. The effort devoted to robotic solutions would be more helpful if it were directed at simple equipment improvements and low-cost robotic devices might provide some useful improvements in safety and cost-effectiveness in the short to medium term. Understanding why "high tech" research efforts have failed, so far, may help to avoid similar mistakes in other ambitious robotics research programmes. Sensors - recent research effortsIt is essential to review recent research results on sensing technology, because a robot which can sense landmines, even using human interpretation if necessary, would be an important step forward. However, there is no evidence yet that a new technique will be available in 23 years to replace the current range of metal detectors. The sensing problem can best be understood in its inverse form. Rather than detecting mines, we need to reliably detect the absence of mines. We must be at least 99.99% sure that there is not a mine in the ground where we are about to step, or the vehicle is about to move. With only a tiny number of exceptions, all anti-personnel mines currently in the ground contain metal components. Sensitive metal detectors can detect all these mines, but also detect thousands of pieces of buried metal junk. The challenge, therefore, is to find additional sensors to safely classify a suspect metal object as "not a mine". Radar (ground penetrating, wideband, arrays, synthetic aperture radar), infra-red and microwave radiometry, explosive vapour sensors, acoustic sensors, electromagnetic induction, magnetometers, and electrical impedance tomography are some of the techniques which have been tried. Vast sums of money have been spent on this research, particularly in military research programmes in the USA, Britain, France and Germany. Trevelyan [16] presents a survey of some recent research results. Ideally, a sensor would react to the presence or absence of explosives. Dogs can smell explosive vapours, but the most sensitive artificial detectors developed so far have orders of magnitude less sensitivity. Unfortunately, dogs are not 100% consistent, and treat their job as a game: they soon become bored. Also, their location accuracy is usually only about 5 metres. A consistent and reliable sensor with the sensitivity of a well-trained dog would be of great value. Several research groups are hoping that sensor fusion techniques will provide major performance improvements. However, the best single sensor alternatives to metal detectors have 80% - 90% detection probability for minimum metal mines which means that the performance has to improve by orders of magnitude before sensor fusion could provide useful results. Experts and Communication ProblemsIn this author's opinion, the main reason why so little useful progress has been achieved lies in human communication failure. Deminers have excellent knowledge of minefields and practical problems in mine clearance. They know less about individual mines, treating all of them with great respect. Military experts have excellent knowledge of explosives, mines and certain types of mine warfare. Sensor experts (researchers and engineers) have excellent knowledge of their own sensing techniques, often in several different application areas. Robotics experts have a broad knowledge of sensors, control systems, mechanisms and computer software. However they need sensor experts to tell them about sensors which are unfamiliar to most robotics people, and usually rely on domestic resident military experts for their knowledge of mines and minefields. While it would be logical to ask them about mines and minefields, they may have little experience of real mine clearance conditions. Perhaps because of this, very few people involved with sensor or robot developments have appropriate working knowledge of minefield conditions. Just one example will suffice. The US Army recently evaluated several metal detectors for work with minimum metal mines (so called plastic mines). The trials were set up by military experts who removed the detonators from the mines to prevent accidents. However, without the metal encased detonators, the mines were much more difficult to detect than they would have been under operational conditions. The electronics experts did not know enough about mines to realise the implications so some metal detectors were incorrectly classified as unsuitable [7]. This illustrates the communication problem. Experience with the robotic sheep shearing project [14] demonstrated the importance of understanding every aspect of the problem, from all possible points of view, to avoid expensive and time-consuming mistakes. Close contact with sensor experts, military personnel and demining groups was therefore essential for the early stages of this research project. By working through the problem with many different people we realized that there were many simple and low cost improvements which would make an impact much quicker than a robotics research programme. Suprisingly, almost no one else seemed to have suggested these. Future Prospects for RobotsSadly, there is little likelihood of success if present approaches are followed. A robotics project requires several important factors to be successful. What is unusual about robotics research is that it requires the successful integration of a number of disparate technologies, most of which must perform at or near ultimate limits. Some of these factors are: Incentive: there must be a large enough economic problem to justify the research expenses. Timescale: the problem will still be there in a few years time. Sensors: the robot must be able to perceive the problem. Cost and availability: the robot solution must be affordable by users. Sensors As we have seen, there are no suitable sensors yet, and available experimental results suggest that there is little likelihood that there will be in the next 2 - 3 years. Incentive Mine clearance programmes are making a big impact and, at current rates of progress, will clear high priority land areas to enable displaced populations to return within a few years. In Afghanistan, for instance, US$75,000,000 over the next three years could achieve this. US$2 billion (the cost of just two major city office towers) would be sufficient to clear most of the urgent land mine problems around the world. Although most of the mines would still be left in the ground, the ones causing civilian casualties and keeping people from their homes and land would have been removed. The rest lie in concentrated areas around military installations, hill tops, and national borders and can be safely fenced off until local authorities decide how to deal with them. The money needed to solve the urgent problem is not large by world standards. Even in Australia, a mining company recently wrote off US$750,000,000 in cost overruns on just one new minerals project, though that was regarded as a serious problem on a local scale! Timescale Significant progress is being made in many different countries, and while the prospects of an effective land mine ban seem remote, there is more awareness now, even among combatants, that land mines should not be used if possible. Even though progress is slow, another decade should see most of the urgent problems solved. Cost and Availability Humanitarian demining organizations already have access to metal detectors and dog teams. While better sensing devices would help, the cost would have to be competitive in the sense that the cost savings in detection would have to pay for the increased price of the detector. However, there are other factors in such decisions other than pure cost alone. External funding for indigenous demining organizations provides a substantial cash boost to the local economy. When confronted with a decision on whether to import cost-saving equipment on a limited budget, they must balance the adverse impact on employment in their community. The objective for demining is to restore land and homes to people so they can become less reliant on external aid funding. If this is at the expense of their economy, then the effort may be counter-productive. Unemployed deminers have to feed their families, and this may mean returning to the armies of local war-lord or drug empires. Any country which develops an effective sensor for detecting land mines and other buried explosive devices will immediately gain a significant military advantage over other countries. Therefore, irrespective of cost, it is unlikely that such technology would be released to non-state organizations. This is not a problem for robots yet. The widespread belief that the global landmine problem can be solved using a combination of advanced robotics, sophisticated sensors, and powerful computing devices is simply a myth. The urgent problem is not large in financial terms. While there are simple and effective ways to solve current problems we should explore those first. The money now being spent on elaborate sensors and robotic solutions is very unlikely to lead to useful improvements. If it were to be spent on simple improvements, the same money could transform humanitarian demining operations almost overnight. Some PossibilitiesIf we take a "lateral thinking" approach, then there are some distant possibilities for robots. While manual demining is the only method that works, it is expensive and time-consuming. The most effective way to reduce costs is to reduce the area which needs demining by being able to declare parcels of land to be free of mines and explosive devices. One way to do this is by chemical sniffing. Sandia Laboratories and some other groups are working on highly sensitive chemical sniffers which could develop into devices to be carried by robots into unknown areas to see whether there are residual traces of explosive vapour. Dogs can also be used in this context - the MEDDS system used in South Africa for instance - see the US Army Website.
References[1] Cain, B. and T. Meidinger (1996). 'The Improved Landmine Detection System.' (EUREL, 1996), pp. 188-192. [2] Daniels, R. A. (1996), Videotape shown at EUREL conference Detecting Abandoned Landmines, Edinburgh, available from ERA plc, Britain. [3] EUREL (1996) Proceedings of EUREL International Conference on the Detection of Abandoned Landmines, IEE Conference Publication 431, London, UK. [4] Janes (1996). Janes Mines and Mine Clearance Techniques, King, C (Ed) Coulsdon (163 Brighton Rd, Coulsdon, Surrey CR5 2NH, UK) Jane's Information Group. [5] Jefferson (1997). Mines, Damn Lies and Statistics, Manchester Guardian, September 1997 (available from http://www.mech.uwa.edu.au/jpt/demining/lies.html). [6] King, C. (1997) Mine Clearance in the Real World. SusDem97: International Workshop on Sustainable Humanitarian Demining, Zagreb, October, pp S2.1-8. [7] King, C. (1996) Personal communication. (Editor of Janes 'Mines and Mine Clearance). [8] McGrath, R. (1994). Landmines, Legacy of Conflict: A Manual for Development Workers, Oxfam, Oxford, UK. [9] McMichael, D.W. (1996) 'Data Fusion for Vehicle-Borne Mine Detection.' (EUREL, 1996), pp. 167-171. [10] Nicoud, J .D. (1995) Proceedings of Workshop on Anti-personnel Mine Detection and Removal WAPM '95, Swiss Federal Institute of Technology Microprocessors and Interfaces Laboratory (EPFL-LAMI), Lausanne, Switzerland. [11] Nicoud, J-D. (1996) 'A Demining Technology Project.' (EUREL, 1996), pp.37-41. [12] Nicoud, J-D. (1997) Vehicles and Robots for Humanitarian Demining, Industrial Robot Vol 24, No. 2, pp 164-168 . [13] Red Cross (1995) Landmines must be stopped: Chapter VI Mine Clearance. Special Brochure, International Committee of Red Cross, Geneva, Switzerland. [14] Trevelyan, J. P. (1992). Robots for Shearing Sheep: Shear Magic. Oxford Science Publications, UK. [15] Trevelyan, J. P. (1996c) 'A suspended device for humanitarian demining.' (EUREL, 1996), pp.42-45. [16] Trevelyan, J. P. (1997a) Robots and Landmines, Industrial Robot Vol 24, No. 2, pp 114-125. [17] Trevelyan, J. (1997b). Modelling minefield clearance statistics. Technical Report, Department of Mechanical and Materials Engineering, University of Western Australia. [18] Trevelyan, J. P. (1997c). Better tools for deminers: International Workshop on Sustainable Humanitarian Demining, Zagreb, October, pp S6.1-12. [19] UNOCHA (1996) Notes on interviews with UN Office for Coordinating
Humanitarian Aid to Afghanistan (UNOCHA), Islamabad. (Available from author).
|
|