My last post was considered more newsworthy than I thought. In a recent IEEE Spectrum article (Hoaloha Robotics Developing Socially Assistive Hardware Platform), Senior Writer, Evan Ackerman made several important observations and comments that I wanted to respond to.
“…but we know is that the robot will likely not include an arm at this time, because there’s no way to add one and still hit Hoaloha’s cost target,…”
This is true and deserves some further explanation. Let’s start with the cost of components. A majority of conventional robot arms include six to seven servo motors to get a similar degree of movement that a human arm has. While you might be able to buy or build such an arm with basic hobby servo motors, to hold up to the load and usage patterns, you really need something of higher quality with good gearing, sufficient torque, and little backlash. Such motors typically will cost in the neighborhood of $1000 each. Then you’ll need to add framing, wiring, control, and other components, so you might be lucky to get by with around $10,000 for a single arm. Further, if you want it to be safe in use around humans (as we do) you’ll need to consider how to add sensing to ensure that as it is moving it doesn’t injure someone or break something. Such a price is somewhat confirmed by the cost of the Rethink Robotics Baxter, that sports two dexterous arms at a price of $22,000. Or try pricing out Barrett Technologies’ WAM arms or any other good quality arm. So as Evan rightly observes we could not easily meet our price target for our robot of $5,000 – $10,000 if the arms are going to cost over that.
While the cost of robot arms may eventually come down that may take some time and might not decrease as rapidly as other components have. Motors are built out of materials like copper and rare earth magnets, both of which are in high demand and not likely to fall in price any time soon. In fact, copper value is such that it is not unusual to hear of thefts from construction sites. An unleased workspace near my office was recently stripped by thieves of all its copper wiring. And you may have noted that the latest minted U.S. pennies now have less copper in them. For robot arms to come down significantly in price, we may need to see some further innovation and there is active research into alternative forms of robotic manipulation.
But there’s more to consider than just the component cost of motors. Motors are electrical devices and so require power. Unlike Baxter that can stay plugged into the wall, our robot is design to be mobile and must carry around its own power supply in the form of batteries. However, batteries also add cost and weight to the overall platform. Consider that we have an embedded PC onboard that roughly has similar processing capabilities as a good lightweight laptop. Just running the processor, display, wi-fi, and other components on a good laptop can drain its battery in around 4 hours. But our robot also has drive motors to get it around and a host of sensors. Adding 7 to 14 motors for an arm or arms to the equation and you likely reduce runtime to about 20 minutes, the rumored time that Honda’s Asimo can run before needing a recharge. This is hardly an acceptable time for being a useful companion. As a result, including conventional robot arms would easily blow our power budget for the robot and likely require recharging so frequently to make it essentially useless.
Finally, there is “cost” in software complexity. Even the impressive Baxter only does specific tasks that it has been explicitly walked through. That’s not a criticism, as it is true for most robot arms. Its creator, Dr. Rod Brooks has stated that it’s design goal was for light industrial, small business use, not a household product as Roomba was. Baxter’s innovation is in its significantly lower price, design for safe use around humans, its ease of programming, and its novel human-interface features as compared to its factory brothers. So far it is not able to generalize its manipulation abilities. Software that could enable a robot to generalize across a wide variety of scenarios is very hard. Some of the best universities are hard at work on it. While children learn to master the coordination between their perception and manipulation at a very early age, it is still very hard for a computer/robot to do. Oh and remember you have to factor safety in the design too. There’s a reason they put cages around factory robots as they have little to no ability to operate safely with humans nearby.
Without a doubt, a robot with arms could provide a lot of value, not only in what the robot could do, but also in interaction (i.e. displaying body language). So while we don’t plan to include arms now, our design easily could accommodate arms when such technology becomes “affordable” in the future. We also are considering alternative ways that enable the robot to fetch and deliver things.
“Without a manipulator, whatever Hoaloha develops is going to have to rely on two things to be successful: user experience and low cost.”
I couldn’t agree more. From the beginning of the company, we have set user interaction and affordability as priorities, essential to our success. To that I would also add the requirement for a useful set of core applications and services as well as safe, reliable autonomous navigation. Great user experience is what enabled Apple to succeed with the iPad and the iPhone, even though not being the first to market with either product category. Both products have not only contributed to their own success, but profoundly affected the competition and the very fabric of the computing industry. In a similar way, iRobot chose simplicity and cost over more sophisticated technologies in their design of the Roomba. For example, instead of opting for Roomba mapping out a room, as the more expensive Electrolux Trilobye did, they went with a simple obstacle avoidance scheme. This enables the Roomba not only to be less expensive, but smaller in profile and thus able to move under furniture. These design decisions contributed to the product’s success as well as a host of imitators. So in both examples, these companies did not rely on dazzling the world with technology, but focused on ease of use and a strong value proposition (benefit at the right price), creating products that users love, reflected in their sales.
“What remains to be seen, however, is whether or not a robot (especially one without a manipulator) has enough to offer to function well as a caretaker replacement, and whether the people who are being taken care of are going to be cool with that.”
There is already substantial, overwhelming data, not only on the reality of the desire and need for technologies to empower seniors to continue to live independently, but also growing research that confirms they would readily accept a robot companion, even more so than their caregivers. Further, because the Baby Boomer segment will fuel a large part of the increase of our aging population, we will have a population of seniors already accustomed to using technology to manage their lives, as well as the care of their families and their parents. This generation has also focused on staying active, healthy, and creative, while meanwhile their children are growing up building robots.
But I must take some issue with Evan’s implication that our objective is “caretaker replacement”. While it is true that we face a shrinking supply of caregivers, and that our goal is to enable seniors to be more independent, it is wrong to imply that this necessarily means we are trying to replace humans with robots, any more than your PC, smartphone, or tablet is a replacement for human interaction. Instead we use these technologies to help us remember important events, stay informed, communicate with others, and even entertain ourselves. In the same way we are designing our robot to be an enabler, a way for people to augment their physical, social, and cognitive limitations. A key part of that is designing our robot to facilitate and simplify communication between users and their caregivers. In a world with a shrinking supply of caregivers, seniors will need not only to be more self-sufficient, but also much more efficient in communicating with an increasingly smaller set of human support.