Webcast From the Wilderness

By Susan E. Cotton

A Texas A&M University computer scientist boasts that his research is better than the cable channel Animal Planet. He says the research is less costly and dangerous but more detailed than the cable channel’s shows: “You can watch Animal Planet, but you don’t have the same involvement.”

The boastful computer scientist is Assistant Professor Dezhen Song. His research concerns robots that can webcast wildlife to documentary filmmakers, natural scientists and the public.

In June 2005, Song and a University of California–Berkeley computer scientist, Professor Ken Goldberg, received $400,000 from the National Science Foundation to fund the research, Collaborative Observatories for Natural Environments. The Texas Engineering Experiment Station administers the funds.

 “You’ve several elements behind ‘collaborative’: collaboration between experts and the general public and collaboration between humans and machines,” he says.
 
In other words, documentary filmmakers and natural scientists will work together with the robots to show wildlife to the public: They will program the robots to recognize a certain kind of wildlife and will set the robots at a certain site. Then the robots will automatically adapt to the site.

“Drop the box there and it starts working,” Song says.

Most sites will lack the electricity required for the robots to work. Sunlight or wind will power the robots’ webcams and wireless networks.

Panasonic Research and Development has donated webcams that are energy-efficient and lightweight. The $1,000 webcams are housed in climate-controlled domes.

Their webcasts will travel via long-range wireless networks to the Web. Then the documentary filmmakers, natural scientists and the public will watch the wildlife webcasts on their computers.

They will continue to transmit their webcasts without further instructions from programmers. The robots will run almost autonomously.

“That’s a key research question: How much autonomy to bring to a system?” he says. “For robots or more generic automation systems, level of automation and how to achieve [it] is always an interesting problem.”

The problem is a familiar one for Song. In 2003 he and Goldberg programmed a webcam, ShareCam, that the public could aim via the Web. They developed algorithms that let ShareCam decide where to aim itself to satisfy the most people when more than one person tried to aim it at one time.

Algorithms will let these robots both decide where to direct their webcams when more than one filmmaker or scientist tries to control them and distinguish among wildlife.

“I think the important part is that we want to know about wildlife,” Song says. “By developing autonomous observatories, it allows us to get close to the animals without being near them.”

Of course, the robots will sit near the wildlife in all kinds of inclement weather — except the cold kind. He and Goldberg had intended to set some robots at sites in Alaska. (“People can sit in their warm homes and watch polar bears,” Song had said.) Yet they learned that the robots cannot withstand such cold weather and instead plan to set the robots at sites in California’s Yosemite National Park in 2006.

In November 2005, Song and Goldberg set a prototype robot, CONE 1.0, at a site in the San Francisco Bay. “This system is the alpha test or beta test of the fully integrated system,” Song says. “We want to make our system more mature before we send it into the wilderness.” Into 2006, CONE 1.0 will webcast birds and boats in the bay — and better than Animal Planet. To see the San Francisco Bay with CONE 1.0, visit http://www.c-o-n-e.org.