Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS
SAF Archives  search ask the scientists in the classroom cool science
scientists from previous shows
cool careers in science
ask the scientists
PREVIOUS SCIENTISTS

Photo of Roy E. Ritzmann Roy E. Ritzmann as seen on Natural Born Robots: Roboroach

Click on Roy's photo to read a brief bio.



q I'm curious to know about how your project is going. What have you and your team discovered and/or accomplished since the footage presented in the show? (Question sent by Jon McMillion)

A Well, making the robot walk is still a work in progress. Many of the problems that the engineers in Roger's lab have had to overcome involve difficulties in controlling pneumatic actuators on a device as complicated as Robot III. But they have overcome many of those problems and redesigned much of the control schemes. The current versions actually are more reminiscent of how the animal's nervous system is actually laid out. So, that is an interesting and hopeful development. Roger is hoping to begin testing the walking control soon. On the biology side, we have just completed a detailed study of the movements involved in climbing over barriers. This has told us quite a bit about the strategies used by the cockroach when faced with barriers that it cannot simply run over using a normal gait. One of my students, Andrew Tryba, has also just completed a study in which he developed a preparation that allows us to study the nervous system of the animal as it is walking in a tether. So, it is doing the correct leg movements, but not going anywhere. This allows us to get even more detailed data on joint movement and also allows us to actually open the animal up and put electrodes directly into individual neurons. These intracellular recordings allow very precise studies of what individual neurons are doing during the behavior. We can also inject dye into the cells and see what they look like. This information allows us to identify the cell as an individual, so we can work out connections between neurons and do repeated studies on the same cells. With this preparation we are about to start asking a lot of different questions. For example we can see how information from higher centers on the size and location of barriers gets ported to the locomotion control circuits that control individual legs and generate modifications for climbing. As we work these kinds of problems out, we can help the engineers to solve these problems in the robot.


q What is the more difficult part of designing and building the robot roach - is it figuring out how the real roach moves, or replicating that movement in your robot design? (Question sent by Shawn)

A Yes, both. Actually for me the most difficult part is figuring out how the roach moves, because I am the biologist on the project and its up to Roger to figure out how the capture the movements in the robot. Since that's his problem, he would probably argue that it is the hardest part. In observing the work on the robot, my impression is that building the robot was not as hard as controlling the leg movements on a freely moving device. That has been the largest challenge for the engineers. Of course the parallel for the biologists is figuring out how the animal's nervous system controls the movements that we documented in the behaviors. This has been an equally difficult challenge and we are just starting to make progress there. If we knew the entire nervous system of the animal really well, we could implement it in control of the robot and the results would be fantastic. In fact, this is why we built the robot so similar to the roach. We can more easily implement the biology of the roach into a cockroach-like robot, and the nervous system and the mechanics of the animal are closely tied. Unfortunately, we will probably never understand the cockroach nervous system completely. So, we need to use other techniques to help fill in the gaps in a logical way that captures both what we know about neural control and what we know about the animal's and the robot's mechanics. Part of that is being done by another member of the team, Randy Beer, who is in computer science and plays a role that is equally important on our team to Roger and me. In fact it was Randy who put the whole team together in the first place. He uses computer techniques such as Genetic Algorithms to fill in those blanks in neural control.

The really fun part about this project is that everyone on the team works so closely together that we can all benefit from each others struggles. Because the legs on the robot were designed to match the animals so closely, the problems that the engineers face and their solutions can give the biologists insight into how the animal has solved these same problems. At the same time, neural control solutions that we find in studying locomotion in the insect can help the engineers solve problems in controlling the robot. So, it's very much a close interaction and it's kind of difficult to say which is harder. Both have inherent difficulties as well as benefits.


q Scientific American Frontiers has a webpage about you which mentions that you study the way insects walk and move. I am interested in similar things, and wondered if you could answer a question: Do most insects immediately have the ability to walk perfectly, or do they go through a "learning-how" phase similar to humans? (Question sent by Zeb)

A Insects know how to walk immediately when they hatch out, so they do not have to go through a learning phase. However, they can learn to position their legs. We have begun to look into questions regarding whether they can learn to overcome problems in walking do to injury. For example, there are two different nerves that control one of the main leg joints. We have lesion one of those nerves and find a deficit in the joint movements that the animal then uses when it walks. So, we are now asking whether it can recover from that deficit by up regulating the remaining motor neuron. People working on cats have done similar experiments and find that they can be trained to overcome such deficits. Other things that we are looking into is to eliminate descending inputs from the brain and see if the lower centers in the ventral nerve cord (the insects version of the spinal cord) can be trained to overcome those deficits and move their legs in a normal fashion. Again, these experiments parallel observations made in cats. Cats that have had a spinal cord injury can be trained to move their hind legs on a treadmill. The advantage of doing this on an insect is that the motor nerve innervation is much simpler to identify. A muscle in a cat may receive many motor neurons and it increases tension by recruiting more and more motor neurons. In an insect the muscle may receive only 2-3 motor neurons, so when you record extracellularly from the muscle, you can be pretty sure exactly what motor neuron your are looking at. So, if we can achieve similar results in the insect that people have found in cats, we might be able to take the projects to new levels that are not possible in mammals.

q What other insects did you consider modeling your robot on, and why did you choose the cockroach? (Question sent by Jonathan)

A We are developing another robot based upon crickets. The goal of that project is to make a robot that is no more than 2 inches in any dimension. Also, the robot must be completely autonomous. That is, EVERYTHING must be on board including the power supply (batteries), control circuits, valves sensors and because it is also a pneumatic device, the air compressor. The obvious problems are that everything has to be very small and light weight and the power consumption has to be kept to a minimum if it is to be able to do any kind of mission. When you miniaturize a robot (or an animal) energy consumption becomes a problem. In order for a small animal to get from point A to point B it must cycle its legs many more times than a large animal. Imagine that you are walking two dogs, one a big Golden Retriever and one a small dachshund (I use this example because I have a Golden and a friend of ours has dachshund mix). The big dog lopes along at a very comfortable pace while the little dog has to move its legs many more times to keep up. The result is that the small dog gets tired much more quickly. So, energy consumption is a problem in small robots also. Some small animals overcome this problem by jumping. For example, crickets and grasshoppers jump to go long distances and then walk to home in on a specific target. So, we decided to model this robot on crickets and have been investigating their jumping ability and righting responses (they don't land very delicately).

This points to a tremendous advantage to using insects for these projects. As the most successful animals on earth, they inhabit just about every niche and lifestyle you can imagine. So, if you keep your eyes open and remember all the insects that you see, when an engineer comes to you with a robotics problem, you can usually find some insect or other arthropod like crabs and lobsters that has solved that problem. The reason for choosing cockroaches for Robot III was that the goal of that robot was to make a robot that would be much more agile than available legged robots. So, we sought to capture the leg movements of a single insect as it easily ran over flat surfaces, inclines and also climbed over barriers. Cockroaches do this VERY well. Also, when we started the project quite a bit was already known about the cockroaches leg movements as it walked, the forces that it developed and the neural control of legs. That information came from a whole lot of laboratories focused on cockroach walking and running. So, we didn't have to start from scratch. Also, I already had considerable experience on cockroaches from other projects that we had done.

One of the animals both Roger and I would like to use for future robots is Preying Mantises. They have a remarkable ability to capture prey with their front legs, then manipulate that and hold it while standing on their remaining four legs. So, it would make a good model for a maintenance type robot. In addition (and probably more importantly), I just think they're cool animals.




 

Scientific American Frontiers
Fall 1990 to Spring 2000
Sponsored by GTE Corporation,
now a part of Verizon Communications Inc.