DSTO & R&D: New tricks from MAGICian | ADM Nov 2010

As the only Australian finalist, Team MAGICian has seen an impressive pairing of universities and industry to develop their robot fleet.

Katherine Ziesing | ADM Nov 2010

Team MAGICian (MAGIC by Innovative Australian Nerds or Intelligent Autonomous Navigators depending on who you talk to) has introduced task-oriented heterogeneity, designing forward, close-contact and rearguard roles and specific Unmanned Ground Vehicles (UGVs) for the challenge.

“To have been chosen from such a competitive international field is testament to the creativity and many hours of hard work of our team of researchers, including students,” Dr David Powers the MAGICian Team Leader at Flinders University and Director of Flinders Artificial Intelligence, Knowledge Discovery and Language Technologies Laboratories said.

“In addition to getting us to the finals, a really important spin-off is that the technology we’re developing is feeding directly into Flinders undergraduate engineering programs and will boost our intake of PhD candidates in the area of robotics.

“We also have ongoing involvement of a group of nine students from Australian Science and Mathematics School who are learning on the job by helping us run all the experiments necessary to refine our robots and produce a winning entry, under the guidance team members.”

The 30kg UGVs based on the Pioneer AT3 employ SICK LMS 111 LIDAR (light detecting and ranging) and Hoyuku laser scanners were profiled last month in the October edition of ADM.

The 10 kilogram Scouts and Protectors based on the HSP Remote-Control Rock Crawler chassis are designed to provide protection for the Pioneers and Disruptors, Scouts scouting ahead and locating the enemy with 360° coverage at five feet per second.

The five-kilogram Rangers based on the Lynxmotion A4WD1 have rearguard and relay functions.

Each of the robot classes are able to be configured for each of the different roles as needed.

The team has found that while each robot was designed for a specific role, they are each capable of performing more than one.

It’s more a matter of determining which robot is best suited to each task.

The family of UGVs are equipped with 802.11 wireless, GPS, IMU/digital compass and Logitech Spheres for close PTZ examination of object of interest (OoI), and in the Disruptor variant the Spheres include a laser designator.

This Sphere configuration was developed, designed and produced by the Flinders University team but more on that later.

900MHz and 3G fallback capabilities are also provided.

The WA Wambots are based on a conventional and heavy, stable robot base and commercial laser scanners, the Pioneer AT3/LMS111/Hoyuku, which they quickly built up into a usable UGV for MAGIC, and they have concentrated on basic navigation and neutralization functions.

“These are expensive and we could only afford a minimal team with two disruptors with four or five sensor vehicles, but this may change,” Powers said.

WA have also built up a minimal bandwidth wireless relay network.

“As team leader I have been particularly impressed at the professionalism and competence of the WA team in ensuring timely and effective integration and testing of successive versions of the systems,” Powers said.

“This is a real testament to the robotics training at University of Western Australia, as well as to the strong array of experienced graduates participating through their industry partners.

“It is important to understand that the concept of robot has moved on from assembly line monsters to embedded applications like the smart cars and autonomous vehicles UWA has developed, both as part of MAGICian and for other projects.

“SA has built up our own design targeted to specific roles at a fraction the cost and weight,” Powers said.

“Our robots all provide continuous 360° vision coverage and the bandwidth to send this to the operators at ground control, where as the WA robots provide a 270° point cloud and can only transmit images slowly on demand by specifically directing the robot and/or a Logitech Sphere Pan Tilt Camera.”

Flinders has also developed a variant of the Sphere that incorporates a laser designator and are being used on the WA Disruptor (as well as currently being used on the SA Crawlers in Disruptor configuration).

This provides a cheap but effective ability to pan/tilt the laser to hit the target and confirm the designation visually.

Flinders has also developed advanced vision algorithms for efficient detection of OoI, as well as accurate visual localization of both UGVs and OoI.

“One of our main areas of focus has been an advanced Human Robot Interface, although our original plans have been curtailed as a result of rule changes and related rulings, and is thus restricted to a far more conventional design than we originally proposed and initiated (which included Speech and Brain Control Interfaces),” Powers said.

“One disappointment is that whereas the original aims of the challenge, and the original rules, encouraged out-of-the-box solutions and radical departure from conventional approaches, the rule changes and rulings have forced a very conventional and less innovative approach to be adopted, rather than revolutionizing defence system interfaces,” Powers said.

“The original rules were fairly general and they [the organisers] have introduced a whole new raft of rulings that disallowed us using our Brain Computer Interface (BCI) where we had the ability to think things or look at images or messages and the computer would know that we had registered them.

“There were also speech interfaces for the robots.

“But we were forced to use a much more traditional approach, using a keyboard and mouse to enter information.”

Dr Powers is at a loss to explain why their technology was not allowed given this is where technology in this area is heading but ADM suspects that it may be for ease of comparison between the finalists and no doubt it would have been a tough call for the organisers to make.

“It doesn’t matter how autonomous a robot is, if the user does not trust it, because whatever level of autonomy you have managed to achieve is useless,” Powers points out.

“We have been doing some very interesting work with DSTO in this area, in terms of cognitive and perception loads.

“What does it take for a person to complete a task in terms of brainpower?

“You want pilots or drivers or whoever focusing on their mission or situational awareness rather than on operating the platform they’re in.

“This is what we’re aiming to understand and demonstrate.”

At the time of writing, two members of the SA team (plus robots) were in WA working on the integration of the two systems, and working on joint tests before the finals this month.