Getting in touch with your surroundings

Citation

Davies, B. (2001), "Getting in touch with your surroundings", Industrial Robot, Vol. 28 No. 1. https://doi.org/10.1108/ir.2001.04928aaa.001

Getting in touch with your surroundings

Getting in touch with your surroundings

Since the initial emergence of the robot on to the industrial and domestic scene, the vision of an autonomous mobile robot has hovered gently above the horizon. The opportunities for such a robot are truly enormous in every imaginable sphere from subsea to outer space. Unfortunately, despite the huge steps that have been made in computer and sensor technology, this hovering image has never landed and become a concrete reality. From its inception, an area that captured the imagination of many people is the application of robots to the domain of personal care – domestic, therapeutic and leisure services. As the needs of the world population change, these activities are steadily rising up the priority scale.

To be able to perform the role of a personal service robot in such unstructured environments as a house, a hospital or even a sports centre, robots must be able to interact with the environment in an intelligent and non-catastrophic manner. Sensors already exist that can provide visual data about the surroundings, data to identify position and data to provide knowledge about other active artefacts within its operating area. Computers are available that are sufficiently powerful to handle these data and turn them into useful information upon which control decisions may be based.

Even so, in the year 2000, all the robots described in the current literature either operate within very controlled and well defined surroundings or require huge reconstruction of the environment to make any sort of autonomous behaviour possible: the Essex Rangers United robodog team referred to in this issue, who I am sure will be playing soon in the premier league, compete on a football pitch which is a very precisely specified zone.

One of the primary requirements of a mobile/autonomous robot is that it can move about in any particular environment without destroying its surroundings. This has led to the perception that "collision avoidance" is of paramont importance for mobile robots. This view is illustrated by the response to a question at the launch of the "Helpmate" hospital robot. The question concerned safety and the possibility of the robot crashing into people, etc. and Joseph Engelberger replied that the robot "carried 15,000 answers" to the question "How do I, the robot, avoid a collision?" With 15,000 possible answers, not only is the wrong answer a distinct possibility, it suggests that the robot is being asked the wrong question. Is it necessary to avoid collisions or could they be tolerated and would "collision control" be a more fruitful approach to solving this problem of navigating our way through an unknown and complex locality?

Human beings operate very successfully in the most appallingly unstructured and cluttered environments. We humans make regular contact with the surroundings – collisions if you like – and not only survive them but learn a great deal about our surroundings from the data gathered during the collison process. These data are processed and stored for future use. For example, if I fail to see a 75mm high brick on the floor and stub my toe on it, I do not keep on kicking it saying "I can't see it, it can't be there!" Nor do I wait until someone else comes along and moves the brick! I add this information to my knowledge base and choose a slightly different route, even if it only means raising my foot further off the ground. Through personal experience, I also make every effort to minimise or eliminate particular potential collisions. Contact with cars, trains, aeroplanes and thrown articles I carefully avoid; in fact, anything that has a size/mass and speed combination that I associate with pain or potential destruction. We all started to learn these critical things at a very early age, gathering data from every contact, inadvertent or otherwise, with the young brain accepting the data generated and starting to practise the art of sensor fusion. As babies we crash into a chair, find it hurts and try to avoid this condition again. We do not need to hit every chair to confirm this, we simply construct a hypothesis that objects of that shape and size are best avoided when we are in motion.

In recognition of this fundamental characteristic of one highly successful autonomous system, is it time for us to be moving on from "collision avoidance" into the realm of "collision acceptance"? Will mobile robots have to tolerate colliding with their surroundings for them to become a reality? Rather than remaining paranoid about crashes and collisions, robots will consider them a part of everyday life and use them as a means of building up their own personal knowledge of the surroundings. No doubt the final path will involve both approaches, and a suitably integrated form of "collision management" mechanism may then be the final solution to the problem.

The gathering of knowledge concerning the operating conditions surrounding an industrial or domestic product already occurs but at a different rate from that observed in humans. For example, the Dyson cleaner has no bag to replace, Jaguar cars emphasise safety through the "intelligent" operation of driver and passenger airbags and most industrial robots highlight the operational sequences and sensor options not previously available to the purchaser. This gathering of knowledge and its incorporation into the next model or product are reflected by the use of advertising terms such as "lessons learned", "intelligent" and "next generation". The company is suggesting that much, if not all, that was experienced during the operation of the earlier machines has been incorporated into this "new" model.

Robots have been with us for a relatively long time. The intelligent sensors, compliant actuators and high speed computers currently available should mean that the fully autonomous personal robot is much closer to being a product awaiting conception than a twinkle in its inventor's eye.

I believe progress will be governed by ability of researchers and engineers to fully and imaginatively utilise existing state-of-the-art technology and fuse the various components into a cohesive product. In particular, the role of "collision avoidance" should change and be expanded and incorporated into an integrated "collision management" system that can tolerate contact with the surroundings. This system would be able to monitor any collisions that occurred and use this information to improve the performance and enhance the behaviour of our own personal robot assistant.

Bruce Davies