Wheels offer speed and versatility when operating on roads, but quickly bog down in off-road terrain. Tank-like tracks improve off-road capability, but cannot move with the speed, agility, and reliability of a military patrol in forests, jungles, or mountains. Simple obstacles, such as fallen trees or large rocks, are quite literally insurmountable for both in any practical combat application.

This led researchers to legged locomotion – more C3PO than R2D2. Various experiments looked at the advantages and disadvantages of multiple legs (akin to a centipede), six legs, four legs, and two. The fewer legs, the more closely a robot can mimic the movements and agility of a human; in addition, equipment and much of the environment is designed for bipeds in a fairly limited range of sizes. However, fewer legs also increase the problems of balance, speed, and the amount of weight the platform could carry – including its own power system.

Mandelbaum was program manager on Big Dog, completed in 2008, and has just begun a second – the Legged Squad Support System (LS3). Each demonstrates both the promise and the extent of difficulty involved, but LS3 is not considered a direct follow-on to Big Dog.

“Before Big Dog, it wasn’t clear we as a community knew how to control a very-high-degree-of-freedom legged platform. Legs can go over any terrain a human can go over, but it is much more difficult to control. When you go off road, it is a much higher degree of variability, which you have to match with a much higher degree of locomotion, then match that with a higher degree of control,” he said.

“Big Dog proved that, but LS3 is looking to carry a real payload – we have suggested 400 pounds – be able to operate at least 24 hours, and cover 20 miles with the same tempo of a regular squad, meaning it must have burst modes. The idea is to accompany a squad and carry a significant portion of their equipment, but not impact where they go.”

The big challenge has always been a self-sustained system that can carry its own fuel source and power system. The bigger the vehicle, the more power required and the larger the engine. Adding a requirement for long endurance adds the additional weight and bulk of carrying its own fuel. DARPA has a variety of programs under way, not all specific to robotics, to develop advanced fuel cells, batteries, and new kinds of engines that might be applicable to a range of robot requirements.

“We still need to deal with the overall power problem, not just the engine, so we’re looking at hybrid systems – internal combustion and batteries, using the batteries when we need quiet operations – sort of a Prius-type engine,” Mandelbaum said, referring to the fourth-generation Toyota Prius, one of the most popular of the new breed of hybrid automobiles.

“Further into the future would be fuel cells; right now, the state of the art is not quite there in terms of the energy and power density required to power something like Big Dog or LS3. However, we have proof in nature that it is possible; all animals are essentially fuel cells. Even if you had to provide specific types of fuel for it, the concept of generating enough power to operate the vehicle from a fuel cell is a great concept.”

DARPA also has looked to nature for concepts of controlled movement to increase efficiency and reduce power requirements.

“When a kangaroo jumps, it extends its tendons and when it lands, those huge leg tendons stretch, which convert their energy into spring energy. By proper timing, it can recover up to 93 percent of the energy from one jump to reuse in the next jump,” Mandelbaum explained. “Our current systems do not recover very much energy, certainly not on the magnitude of a kangaroo, because the state of the art in control and coordination of motion is not there.