Hitting the track with Boston Dynamics’ Handle

Boston Dynamics recently unveiled its latest creation, a two-wheeled robot named Handle built to hurdle obstacles at speed. Claire Apthorp reports on this leap in robotics and considers whether, with the company reportedly up for sale, the designs will ever make it into development for military customers.

In January, footage appeared online of a bi-peddled two-wheeled robot known as Handle, built by Boston Dynamics. The video, which shows a large, humanoid type wheeled robot moving dynamically at speed, jumping, spinning and continuously rebalancing itself, was accompanied by audio from Boston Dynamics founder Marc Railbert referring to the robot as ‘nightmare inducing’.

Handle is the company’s first foray into wheeled, as opposed to legged, robots.  “This is an experiment in combining wheels with legs with a very dynamic system that is balancing itself all the time,” Railbert says. “It has a lot of knowledge of how to throw its weight around, which it uses to help stabilise itself.

”This is much more efficient than a legged robot. It can carry a reasonably heavy load on a small footprint, and it’s basically an exercise in seeing if we can do something like the humanoid that has less degrees of freedom – [and] eventually could be less expensive – but still have that significant capability.”

Although the footage caused a stir across the unmanned ground vehicle technology world, a mere three months later Boston Dynamics’ parent company, Google, is already distancing itself from the robot.

Rags to riches

Boston Dynamics was acquired by Google in 2013 as part of a wider flurry of purchases that included multiple robotics companies. At the time, Boston Dynamics was best known for a number of US military-backed projects including Atlas, BigDog and Cheetah.

Atlas was the leading project, a 6’2”, 330-pound humanoid robot developed for the US Defense Advanced Research Projects Agency (DARPA). It made its debut in 2013 at the DARPA Robotics Challenge.

"Atlas was the leading project, a 6’2”, 330-pound humanoid robot ."

At the time, Atlas was one of the most advanced humanoid robots ever built; essentially a physical shell for the software brains and nerves to be developed by qualifying teams for the challenge. In the 2013 challenge the teams guided the robot through a series of physical tasks representative of what might be encountered in disaster zones; by the 2015 challenge the focus had switched to developing on-board power and wireless communications. Its limbs, already capable of a range of natural movements, were strengthened, and its durability and appearance were improved.

Most recently on this programme, a new version of Atlas designed to operate outdoors and inside buildings was introduced in February 2016. This smaller (5'9" tall) and lighter (180lbs) version is designed for mobile manipulation, and is electrically powered and hydraulically actuated. It has sensors in its body and legs for balance and LIDAR and stereo sensors in its head to avoid obstacles, assess the terrain, help with navigation and manipulate objects.

Pack mule

BigDog was also developed with the backing of DARPA. The rough-terrain robot, powered by an engine that drives a hydraulic actuation system, has four articulated legs like an animal with compliant elements to absorb shock and recycle energy from one step to the next. Standing at 2.5ft tall, 3ft wide and weighing in at 240lbs, the system had the ability to run at 4mph, traverse steep, rough terrain and carry a 340lbs payload. Later work saw a manipulator arm added to make it capable of dynamic manipulation under funding by the US Army Research Lab.

BigDog helped inform research in perception, autonomy and mobility that culminated in DARPA’s Legged Squad Support System (LS3) for the US Marine Corps. This system was carried out as part of work to reduce the burden on the dismounted solider through the development of a highly mobile, semi-autonomous robot able to carry 400lbs of kit, integrate with a squad of soldiers or marines in a “natural way, similar to a trained animal and its handler”.

This pack mule robot concept had a number of goals: in addition to carrying a significant proportion of the squad’s load without hindering the squad’s mission, it could serve as a mobile auxiliary power source to allow troops to recharge batteries for radios and handheld devices while on patrol. A number of these key capabilities were demonstrated throughout the programme, including advanced mobility and exercising its perception capabilities. The system was able to carry a full load and enough fuel for a 20 mile mission lasting 24 hours. It automatically followed its leader using computer vision and was able to travel to pre-set locations using terrain sensing and GPS. However, the marines ultimately lost interest in the project by the end of 2015 due to the associated noise levels of the engine, deeming it too noisy for combat use. Both the LS3 and BigDog were consequently shelved.

Somewhat similar in design, Cheetah was also a four-legged robot with an articulated back that flexed back and forth on each step, allowing it to run at faster than 29mph. This project was funded by DARPA’s Maximum Mobility and Manipulation (M3), which sought to significantly improve robot capabilities through fundamentally new approaches to the engineering of better design tools, fabrication methods, and control algorithms.

To market, to market?

Despite these projects causing significant excitement across the robotic technology world, all is not well for Boston Dynamics. Google’s parent company Alphabet has been distancing itself from the technology, with media reports suggesting that the fear-factor involved with robotics developing (arguably creepy) human-animal likeness is not something it wishes to be involved in. Amidst reports that Boston Dynamics has failed to fit in to Alphabet’s wider corporate culture, the company is understood to be up for sale with questions being raised over the marketability of its products.

It is a fair question. While development agencies such as DARPA will fund and get excited about the very leading edge of robotic technology, a small fraction of its projects ever see the light of day in a fully funded development project, and even fewer make it into production. So while DARPA has seemingly embraced this direction of technology development, it bears little resemblance to the real-life requirements of the US military with regards to where it is heading with unmanned technology.

There is no doubt that some of the technology that has emerged from Boston Dynamics over the past decade has pushed the boundaries of what was ever thought possible from unmanned and robotic systems, but the company may find that it has walked the line between functionality/applicability and gimmick a little too closely.