December 29th, 2012

Transcript – wambots

ROBOTS:

We, here at the University of Western Australia, in sunny Perth, are speaking to Dr. Thomas Bräunl. He heads the robotics and automation lab.

Professor Thomas Bräunl:

My name is Thomas Bräunl, I’m professor for computer engineering at the University of Western Australia. I’m directing the robotics division up here at the University. I have done numerous robotics projects over the last 15 years here, and recently I have also gone into the area of electric vehicles.

ROBOTS:

And the reason I’m here is to speak to you about the Wambot project. But could I get a little bit of an idea of the regional challenge that you have faced a couple of years ago with the project?

Professor Thomas Bräunl:

The Wambot project came from the initiative of the Magic 2010 National Robot Challenge. We had been working before that a number of years with small mobile robots. We also had some medium size platforms big enough to carry a laptop system. Then with the Magic 2010 challenge we had the opportunity to build a new robot team. So we chose a commercial robot base equipped with a number of high quality sensor systems, an automotive PC to drive the system, and then basically looked at what is available in the public domain software to get us started, and then carry our solutions around that.

ROBOTS:

The original competition was in 2010?

Professor Thomas Bräunl:

Yeah like the competition name Magic 2010.The initiative I think was in 2009, then the rules were sort of specified early 2010. There was a preliminary round at each of the participant’s home turf so to speak, and then there was the final towards the end of the year late of 2010. The challenge itself comprised a number of different phases and involved a number of different tasks. But the main concept, the main idea was basically having an autonomous team of mobile robots. Not just a single robot, it has a to be a team or a swarm of mobile robots and they should explore an unknown environment, then share the data. Basically the individual smaller maps generated by  each of the robots have to be shared and then put together into one continuous resulting map. They also have to identify different objects and different people but that was mostly color coded, so relatively simplified.

ROBOTS:

The result of the competition, your team came (…)

Professor Thomas Bräunl:

Yeah we made it to the finals, which we were very happy about. We were the only Australian team in the finals. We came in an equal fourth in the final.  The 3 winners were the 3 US teams; they have been doing exactly that sort of rigid scenario a number of years before. They already had their robots when they started. We were to create the robots, build them as part of the challenge. We were always hoping that there would be a second challenge with identical or similar task to complete but unfortunately it didn’t happen. Because the first time you go into a competition you learn so much, I think the second time you do it, you are much better prepared for it.

ROBOTS:

I see. So there has not been any talk of a second competition?

Professor Thomas Bräunl:

Yeah there was a talk to repeat this; the organizers originally said not the following year but maybe 2012. That did not happen, this year there was a new project being announced for 2013 but that is going to be completely different to the scenario we had before. It is more in the rescue and humanoid area so something completely different to that original challenge.

ROBOTS:

About the systems that you used to build your machine: you used a lot of off-the-shelf hardware to do that. So the real challenge is in software. How you went around about doing that?

Professor Thomas Bräunl:

Yeah. I think the real challenge is mostly in software. That’s depending on what your approach is. You can build your own hardware but that would probably be taking at least a year until you have everything up and running. So it’s probably more economical if you do not have the resources or make this your main area of research to buy a robot base, then equip the base with the sensors you want to use and then concentrate on software, and that’s a challenge in itself.

ROBOTS:

Were you using processors that you used in other projects for this project or you did say automotive PC?

Professor Thomas Bräunl:

Well the choice of the automotive PC was mainly to have something that’s ruggedized and runs on 12 Volt. We have a similar PC for one of our automotive projects so there’s some synergy there. But beyond that, it is completely separate. We had some previous experiences with similar basis that we had previously for teaching purposes. We used the Pioneer basis for that competition. We bought different versions of the Pioneer basis but we had some experiences with them so we knew how to interface them, we knew the software so it made it easier for us to get started.

ROBOTS:

As far as the operating system that you used: an off-the-shelf operating system or Linux or (…)?

Professor Thomas Bräunl:

The robot itself has a small operating system or firmware that comes with the robot itself on the car PC from the manufacturer. We had a long discussion on what to use. In the end, we let the team decide because the students working on the project have to be happy with i, so we ended up using Windows XP. For a number of reasons it was a straightforward simple decision. For a number of other reasons it was not such a good decision, because of things like having to battle viruses a few days before the competition, that sort of thing. At the moment we have new group of students working on the robots re-implementing everything. We are now using ROS as a basis, so we have linux installed on the systems, creating something that is hopefully more stable and can be used for further projects.

ROBOTS:

As far as the way you are scanning and mapping space: you are using laser, scanning and video camera or (…)?

Professor Thomas Bräunl:

Yeah we have a number of sensors. So we have 2 laser scanners, a GPS, vision and an IMU, plus we have the robot’s own sensors, especially wheel shaft encoders to get its position. We ended up just using the IMU for localization; GPS was not really accurate enough. It was okay getting points out of those but you had too much of a drift obviously it would not work indoors. The camera was only used for detecting objects and people. So for the mapping we were basically just using the laser scanner and the IMU for positioning and that worked quite well. And one of the PhD students, Ronald Reed‘s project was exactly in fusing multiple maps so we could use this ideally for this project because that is exactly where we need this. We have individual robots that all have local maps and then we have to use some statistical method to bring all these maps together because there is always inaccuracies. You have positioning error. You have rotational error so the whole SLAM architecture has to be extended to multi robots SLAM.

ROBOTS:

Do you use swarming algorithms at all?

Professor Thomas Bräunl:

We had little bit of AI component but since that competition did not allow you to interfere briefly from an operator console, we did not go much further into this. Ideally in a fine scenario there should be some small intelligence so the robots find their own positions and negotiate amongst themselves where each robot is going. For the actual competition we placed markers on the control console so the robots were given initial points where they should be going and then further explore from there.

ROBOTS:

As far as the winners of the competition versus your team: did you get any lessons learnt? Were they able to exchange data to give you an idea of what advantage they had or disadvantage?

Professor Thomas Bräunl:

Yeah we exchanged some data with them. We also exchanged some students, some of those students went to the US for a few months to exchange ideas and get some new knowledge. That worked out quite well. We also got measurement data from the robots from the competition and could use them in our own algorithms to see how that works. That was quite useful and quite a good cooperation there.

ROBOTS:

Where is the team headed now without the competition?

Professor Thomas Bräunl:

Yeah without the competition it is always difficult to get the motivation going. That’s, I believe the one advantage about having a competition. You have a deadline, you have a certain goal, so you have the motivation, you have a task that you need to accomplish. It makes it easier I think. At the moment we have about 4 students working on this project. Before we had much, much more. I think it was, depending who you count, between maybe 15-20 students working on the project.  So now it is a much smaller number and without this fixed deadline things get dragged out so it is not so easy to accomplish things in a short time frame. But I think we are making good progress trying to get a solid basis not just for this particular task but also for other robotics task, which we want to achieve with the robot basis that we have now.

ROBOTS:

What do you see as the future as far as where robotics is actually going?

Professor Thomas Bräunl:

Yeah this is always a very difficult question. I mean you have to distinguish robotics as in stationary manipulators, which is always the success story of robotics.  Every large manufacturing company uses robot manipulators, especially in this automotive industry. With mobile robots it is not so much a success story. There is no real industry, there is no real application, if you don’t count maybe cleaning robots and lawn mowing robots, which is maybe a small industry there. So without the big industrial application it basically remains a research area which has lots of potential, but it really does not enjoy the industrial funding and industrial development that is required to bring this to a higher level.

ROBOTS:

I looked at your website: you were looking at underwater robotics and you were talking about a pipeline tracking robot? Was that simulation or a (…)

Professor Thomas Bräunl:

It was both. So we have built a simulation system for underwater robots that is called Sub sim it’s available for free from our website. We have also made this open source now. So whoever has an interest in this one, can download this. The story about this one was: we were offered some funding for creating a competition similar to the US AUVSI competition for underwater vehicles that would allow Australia and the neighbouring region to have a local competition. Because if we have to go to the US, it is very expensive because of airfares and the shipping of robots over there. The idea was that Australia has a marine industry, thus it would be nice to have a local competition for students doing something with autonomous underwater vehicles.

So we developed that simulation system basically as an avenue for universities to compete but not having to build real hardware. The second part of the project was to have a unified underwater platform. We have built it as well so we have 2 models of underwater robot that could have been used as a standard AUV for that competition. Of course you would have universities building UAVs for that particular purpose. Unfortunately we only got to the development stage and there was no further funding there. The person who had championed this project left our sponsor and unfortunately it did not get picked up so we are still without an Australasian competition in this area.

ROBOTS:

The other thing I noticed: you were talking about the (…)?

Professor Thomas Bräunl:

Yeah that was an interesting story as well. We partnered with a Canadian university. My friend called Jacky Baltes had the original idea. He’s from the University of Manitoba in Winnipeg. The idea is we would like to participate in the grand challenge similar to the autonomous car competition that happened a few years ago. But we do not have the same research funding to be able to do this. You would need basically a million dollar budget to get something like this going. So the idea was we could do a very similar research just on a smaller scale. So instead of having a real hummer car equipped with all the sensors and having the ability to drive autonomously, we basically built some model hammers of about half a metre long, put a linux PC in them, stereo cameras, some additional distance sensors and then tried to achieve sort of similar goals which is basically waypointing but also object detection and obstacle avoidance and drive this thing for example in the cricket oval(?). All that a lot cheaper and trying to achieve some similar goals as if having a big budget and a real vehicle in that competition.

ROBOTS:

Did you get any tankers?

Professor Thomas Bräunl:

No unfortunately not. We have done this in Canada and here in Australia but we did not get any other universities joining. It was meant to be an open project and see whether some others would take this up. But unfortunately at this stage it was only the 3 universities involved.

ROBOTS:

The kind of students that you have: if you can give an idea from where you are coming from?

Professor Thomas Bräunl:

I mean my personal background is computer engineering but we have students from many different engineering areas. We have mechanical engineering students, mechatronic students, electrical engineering students, computer engineering students and computer science students. The problem I am seeing now and sort of the recent curriculum development especially here is that programming is not really that much valued in engineering degrees outside of computer science and computer engineering. So we have come to a situation where, especially electrical engineering students, unless they have some extracurricular activities basically cannot program. And although many have interest in robotics, the majority of the work is always in software design.

So we come to the situation where we have students who have interest but we cannot give them certain projects. Because they do not have the necessary background, they cannot program, there is not enough time within one year to start learning how to program and being able to produce software at a level we require so they basically cannot do the project. I think they are missing out on a very valuable skill because whether they work in the industry or for most of the jobs that we were doing, it is expected that they are able to do a certain level of programming as well. So I think it is essential that that should be done for all the engineering disciplines.

ROBOTS:

And thank you Dr. Bräunl. I spoke to one of the final year students Kellan who is currently working on a new Wambot project.

Calum:

I am Calum Meiklejohn. I am a final year student at UWA, I am studying Mechatronics engineering and I have been working on my final year project on Wambots which are basically just a small four wheeled robot equipped with some laser range finders. We have got 5 of them and the goal is to get them to drive around the area exploring autonomously and creating a global map. So we entered them in a competition in 2010. Our uni came 4th and the task was to explore an area of 500m by 500m. The robots are equipped with a whole range of other sensors as well: GPS, another small laser range finder, odometer.

ROBOTS:

The competition is over. And so you’re really looking at the problems from a different perspective.

Calum:

So the competition used software, which was not free and open sourced. So when that finished, a whole bunch of companies took back their software. And we were left with these robots, which did not work. So the goal is now to get these robots operating on the ROS operating system which is a robot operating system and it contains a lot of packages and programs. It is a good way to set a robot up. So I have been trying to port a lot of the current from the competition unto the ROS system as well as just getting them up and running which I am hoping to do pretty soon.

ROBOTS:

So this is really an enjoyable task?

Calum:

Yeah I have had a lot of fun really. It is fun driving a lot around to actually be seen doing something.

ROBOTS:

I hope so. So you are very far away from getting to your goal or do you have set time to get this out? Is there a deadline?

Calum:

Yeah, deadlines are on the 22nd of November so I have to hand in my draft thesis so that would be the very final deadline. But I am hoping within a week I would be able to get them actually exploring. We have already got the whole map building software working, navigation works and we can send it a goal. It can drive up to that position, avoiding obstacles even moving obstacles like these people and so the last part is just writing a program that selects parts in this map which look good to go and explore.

ROBOTS:

Okay. Well thanks very much for that.

| More