Artificial Autonomy in the Natural World: Building a Robot ... - CiteSeerX

0 downloads 0 Views 367KB Size Report
will live free on agricultural land, hunting and catching slugs, and fermenting the corpses to ... and 'digesting' organic material. Such natural ... problems of predation are much more challenging than grazing on plants. However, the prey species should be reasonably plentiful and not require rapid pursuit, which would be ...
Artificial Autonomy in the Natural World: Building a Robot Predator 1

1

1

2

Ian Kelly , Owen Holland , Martin Scull , and David McFarland . 1

Intelligent Autonomous Systems Engineering Laboratory, Faculty of Engineering, University of the West of England, Bristol, BS16 1QY, England [email protected] 2 Department of Zoology, University of Oxford, England

Abstract. With a few exceptions, today’s mobile robots, however complex, are not truly autonomous. At some time, they all require humans to supply them with energy and/or information; most also require other forms of assistance. In complete contrast, even the simplest animals are totally self-sufficient. We describe a current project1 which aims to construct autonomous robots with animal-like self-sufficiency both in terms of energy and information. The robots will live free on agricultural land, hunting and catching slugs, and fermenting the corpses to produce biogas, which will fuel the generator providing the robots with power.

1 Introduction During the last two decades much research has been carried out into the design and control of so-called autonomous robots. However, most of these robots still require some intervention from humans in order to carry out their task(s). Forms of human intervention include supplying information and energy, physically assisting the robot, and modifying the environment to suit the robot(s). For robots to be truly autonomous they would have to be able to carry out their entire mission without human intervention. There are of course a few examples of robots which achieve a high degree of autonomy, in that they carry enough fuel for their mission or can use radiant energy from their environment, and can control themselves without human intervention. Examples include missiles, smart torpedoes, and some spacecraft. In addition some automated cleaning and materials handling AGVs use opportunity battery charging to achieve a degree of autonomy. But while we might be inclined to congratulate ourselves on our achievements in improving the autonomy of robots, we must also recognise that even the simplest animals exhibit a degree of self-sufficiency and independence which is immeasurably superior to that of the best of our robots. This project represents an attempt to design and construct a robot system, with energetic and computational autonomy comparable to an animal system (as [10] urges). In order to make the mission a real technical challenge, we decided that the system would have to obtain its energy in the same way as most animals - by finding and 'digesting' organic material. Such natural resources are found in particular types 1

This project is supported by an award from the EPSRC ROPA programme

of places, and are destroyed by being used; the process of foraging for food must therefore deal with the issue of where and when to look for food, when to revisit an area where food was previously found, when to abandon a site which is not producing food, and so on. Of course, the organic resources, or food, must be converted to a form of energy that the robot system can use. We propose to convert the organic material to electricity by first fermenting it to obtain bio-gas, and then using this biogas to power an internal combustion engine driving a generator. Using modern engine management techniques, it is possible to use bio-gas with as little as 25% methane, and to recover up to 45% of its energy. The organic energy source should be a mobile animal species, since the technical problems of predation are much more challenging than grazing on plants. However, the prey species should be reasonably plentiful and not require rapid pursuit, which would be difficult to achieve at a reasonable energy cost. Since there would be a certain energy cost associated with finding, catching, and consuming any creature, however large or small, the prey species should be reasonably large, to give a reasonable margin over energy expended. Finally, it should preferably be a pest species subject to aggressive control measures, so that the system could be perceived as doing something of actual use that would have to be done anyway. All of the above criteria are met by the slugs found on agricultural land, especially Deroceras reticulatum [8]. They are slow moving, plentiful, large, and destructive - UK farmers spend over £20m per annum on molluscicides and spreading them [1]. Slugs are also potentially more suitable for fermentation than some other possible target species: they do not have a hard shell or exoskeleton and have a high moisture content. The agricultural ground where slugs are pests usually takes the form of a well cultivated seedbed containing winter wheat or potatoes [2]. Such ground is soft - so soft that moving a heavy fermentation vessel over it would consume large amounts of energy. The fermentation vessel/gas engine/generator system will therefore be fixed, and the robots will deliver slugs to it, and collect power from it rather like some social insect colonies. Because the fermenter will require a certain amount of energy to cover its operating overhead, it seems extremely unlikely that a single robot would be able to gather enough energy to service both its own and the fermenter’s requirements; we will therefore require a multiple robot system. This confers other potential advantages: some search tasks are more efficiently performed by a number of communicating robots than by a single robot; perhaps more importantly, the multiple approach gives a potential for achieving reliability through redundancy. This paper describes the progress made to date in the design and construction of the robots, work on the fermenter is to begin shortly.

2 The Robots The key features of the robots are; that they must be energy efficient, operate in unmodified agricultural fields, and be protected from the weather, slime, and mud. To minimise energy usage whilst foraging each robot will scan the ground and catch slugs using a sensor and gripper mounted on a long, light articulated arm - the energy required to move this arm is much less than that required to move the whole robot over rough ground. The optimal arm length is a function of the distribution of slugs and off the power required to operate arms of different lengths in relation to the power

required to move the robot. Our calculations showed that a 1.5m long arm would be the most energy efficient. The other requirements of the arm are that it should be light, stiff, easily controllable, and capable of moving in all directions; it should also have a fairly simple construction, increasing its reliability, making it easier to manufacture and reducing its cost. A design consisting of two 0.75m tubular sections, with a hinged joint between them, was chosen (see Fig. 1). To allow the arm to rotate around the whole robot it is mounted on a turntable located in the centre of the robot’s chassis. The chassis is large enough to maintain stability in all directions when the arm is fully extended. To meet the requirements of lightness and stiffness the arm is constructed from aircraft grade carbon fibre tube. To keep the arm structure light, the motor and gearbox required to provide movement at the elbow joint are mounted on the turntable; a lightweight toothed belt inside the arm transmits the drive to the elbow joint. Since the numbers of slugs on the surface peak in the early evening and just before dawn, the rate of gathering them during these periods must be as high as possible. To this end the arm motors and gearboxes were selected so that the arm can move from fully retracted to fully extended or vice-versa in under 1.5 seconds. Self locking worm gearboxes provide the required motor speed reduction, and allow the arm to be held in position without consuming any energy.

Fig. 1. Prototype three fingered gripper with wiper blades and compliance gimbal (left), and the arm and gripper system mounted on a turn table (right) The arm’s end-effector is a robust lightweight gripper that is able to pick up and release both wet and dry slugs, regardless of their size and orientation, and any o irregularities in the substrate. The current design consists of three fingers at 120 spacing, operated by a single miniature motor. As the fingers close, they meet underneath the slug so that it can be lifted; wiper blades ensure the slug’s release when the gripper is opened. Slugs are detected and targeted by a camera mounted in the centre of the gripper, away from slugs and mud. To allow for contour matching with the ground, and to ensure that the view from the camera is always perpendicular to the ground, regardless of the arm’s extension, the whole gripper assembly hangs freely on a gimbal. When scanning for slugs it is possible to lock the gripper assembly, thus stopping it swinging, by fully opening the gripper. Each of the three wiper blades has a plate on the end to allow for passive alignment with the contour of the ground, ensuring that all three blades move under the slug when the gripper closes. The gripper’s mechanism will be protected from the weather and mud by a flexible rubber cover. Locating the static fermentation station in a large muddy field, where wheel slip will be inevitable, will be achieved by using a combination of the Differential Global

Positioning Satellite (DGPS) system, and an active infrared localisation system [4], [5]. DGPS can also be used for mapping the locations of grazed areas, so that good patches can be found again, and over-grazing can be avoided. (This last point may not be a problem: a study allied to this project [3] found that removing all surface slugs from a location every few days does not, in the medium term, appear to reduce the number of available surface slugs. This is thought to be because there is a large reservoir of underground slugs which can rapidly replace those that are removed.)

3 Sensing and the control strategy A system is required that can successfully detect slugs, under sparse vegetation, against a background of rough earth. This task could potentially be achieved by several different types of sensors; we have opted for a vision based system since it offers the best combination of size, weight, cost and effectiveness. We have opted for a monochrome CMOS image sensor that is lightweight, low power (