In the days after Colin introduced the AVA “concept robot” (as defined by Gerry Caron, director of product management for iRobot), the buzz on the AVA has been superb. Since announcement, the traffic on iRobot and the iRobot AVA has been phenomenal.
But, as reported by MSNBC, Caron states:
“We want to show people a vision of what they can have” at some point with a home robot that’s designed for several uses, Caron said. (Tell them, AVA: “No, I don’t do toilets.”)
And, while the features discussed are quite impressive (as I mention in this post on the AVA), I see some serious design concerns on the device as an effective “presence system”. In this post, let me discuss them further.
UPDATE (1/13): just found BotJunkie’s video on the AVA and a conversation with Mark from iRobot where he clearly states that the AVA is a “technology demonstrator” not a telepresence robot. The idea (from what I gather) is that AVA is showing what is possible from iRobot for discussion sake – and did not want to handicap discussions with constraints for telepresence.
CONCERN 1: Drive Mechanism / Clearance
In all of the videos I’ve seen of the AVA, she moves quite gracefully from one end of the pen to the other.
And with its holonomic base (read: can move in any direction equally), she moves quite fluidly. But, my concern is with AVA’s omnidirectional wheels that are in the tripod base.While omnidirectional wheels provide excellent movement, movement on non-flat surfaces (like a shag carpet) or any form of incline will be quite a challenge. And, as seen in the BotJunkie video below (about the 3:40 mark), the shaky nature of the mecanum-wheels as the AVA moves may cause pilot concerns as the cameras shake with the high-frequency of the wheels vibration.
Floor clearance is another concern (see the distance from the edge of the bottom bumper to the floor); for if there is anything in the way of the AVA, it may have difficulty moving over it.
CONCERN 2: Gaze Issues
While I was amused by the iPad showing off the Scoomba cleaning floors and other nice video snippets, I noted two major concerns.
- iPad is not hooked up at all. Checking this shot from Engadget, you can see that both the power cord and the Ethernet cable are not even hooked up. While I love that the “concept” is presented, I think there is a bit of time going before the AVA is controlled by the iPad (or Android tablet).
UPDATE: in the BotJunkie video, Mark explains that the system is “head agnostic” – allowing for any tablet to be placed on the head of the mobile base. So, as a technology demonstration, I could see how this is not as important.
- Gaze is poorly matched. If you note on the photo to the right, the camera for the pilot to see from is currently below the iPad (which I suppose would be the “head” of the pilot). Thus, the system still suffers from the gaze issue – not ensuring an effective presence for the pilot or the participants.
CONCERN 3: 3D mapping system
While I can not speak to how effective the SLAM/mapping system is with a low-cost LIDAR, sonar, two PrimeSense components that the AVA has, it will be interesting since, from the photos, it seems as though the LIDAR (laser scanner) is positioned at the floor (see the dark components at the front of the AVA base).
UPDATE (1/14): I got an email telling me that one of the PrimeSense cameras is pointing at a 45-degree angle from the head (look at the neck and angle 45 away from the neck) toward the floor which gives a much better forward view of what is in front of the AVA. Much better to know the PrimeSense cameras are being used in an effective manner.
Other Things
Since the AVA is a “concept” robot, I could complain about the sound system (e.g., using the iPad’s anemic speakers) or no perceptible microphones should not be big issues to complain about. Additionally, as a pilot, I am not sure I would want other people changing my physical height by just a touch of a bumper. But, as I said, AVA is a “concept” robot technology demonstration, not a product yet.
See the video below (from BotJunkie) for more information:
Don’t you like anything on the AVA?
Actually, I love a number of things from the AVA. My favorites are:
- The fast rotating head – the fact that head movement can better emulate our natural head movements excited me.
- Extending neck – again, another one of the features I think are critical in establishing presence and intimacy with a particpant
- Concept of SDK for developing robot apps – I love that iRobot is looking into that through the iPad and Android system – I will be quite curious how the development of APIs between these OSes and the iRobot OS comes about. As suggested in the MSNBC article, I think the goal for showing the AVA to the market at CES 2011 was to recruit some programmers that are closet robotists and convince them to get over the hump (which I love the idea).
So – what I take from iRobot’s entry into the space is:
- With Anybots and vGo getting a lionshare of the telepresence robots discussion, iRobot had to make sure it got some market visibility as well.
- Since their primary space seems to be military and vacuum cleaners, allowing their developers to explore other options is good for company/engineering morale (see Google and Facebook’s 20% time story). And, according to some lore, the telepresence robot was iRobot’s first product idea back when it started (but i have no solid confirmation of this).
UPDATE: in listening to the BotJunkie video, I must give credit iRobot with the fact that they wanted to use the AVA as a jumping off point for ideas. I truly understand how, when discussing robotic applications with individuals, it is not as easy as drawing wireframes for websites to visualize what is possible.
[Image Credits: all from Engadget article: iRobot AVA Chills With Us at CES]
Hey Sanford, I thought I might share a few observations:
— You’re spot-on with respect to holonomic bases using mecanum wheels (eg. Segway RMP). They have problems with thick carpet, cords, and even small obstacles like doorway thresholds. However, I presume that a powered-caster design (eg. PR2 or Nomad) could be substituted, which generally do not suffer from these problems. Or, I suppose you could just go with a variant of differential drive to keep costs down. What are your preferences?
— It sort of makes sense that the iPad isn’t connected to the robot. It probably communicates via WiFi, which has the added benefit that you could control it from a remote computer (phone, tablet, PC) as well. Incorporating a power cord should be fairly straight forward.
— I presume that applications would be afforded certain lockouts — like preventing touch interactions from changing the robot height while being teleoperated.
— As for LIDAR, it seems that it is more of a backup sensing modality. Presumably the Primesense depth camera could provide all the sensing for mapping, obstacle avoidance, and human-robot interaction. However, I’ve heard rumblings that backup sensing modalities may be a required robot safety provision.
Hey Travis – thanks for the feedback. I should say that I have updated this article after I found BotJunkies video from CES which was extremely illuminating and well done (way to go Evan).
The iRobot rep was very clear that this was a technology demo, not a prototype for product development. My assumption is that the engineers know a bunch of these concerns already.
Re: drive mechanism – IMHO, I would more than likely go with the diff drive solution, for both cost and power reasons. But I understand the allure for the holonomic base and the desire for an effective turning radius. One of the reasons why the Texai was designed the way it was – making sure that the turning was not onerous. So – the fun part will be applying these concepts (e.g.; powered caster, diff drive) and choosing the tradeoffs.
Re: the iPad/tablet interface – I worry that while a person can develop an interface for the iPad (as in a touch control – like a traveling kiosk); as an interface for telepresence, it concerns me. And since the iPad/iRobot interface is yet to be clarified, I worry that the “ease of development” idea (e.g., “I can program an iPad, which means I can program a robot”) may not be as easy as the prior articles have make out.
Re: feature lockouts – I agree. After watching the BotJunkie video, I now see that it was a demo model, not a particularly designed for the telepresence application. I have spent all the time looking at the systems from a “presence’ lens; I must be careful not to be as harsh.
Re: LIDAR being placed for the safety provision you speak of – I think you are the visionary who sees that, and I know iRobot will explain its rationale (more than likely with your hint…)
One more great site with video: http://hackedgadgets.com/2011/01/14/irobot-ava-mobile-robotics-platform-at-ces-2011/?utm_source=twitterfeed&utm_medium=twitter&utm_campaign=Feed:+HackedGadgets+(Hacked+Gadgets)
Much happier to hear the engineer who described the AVA as a “concept car” showing off their technologies. I think the early press were trying hard to place the AVA in the space of the Anybots, vGo, Intouch and I can see that it is a demo model. Nice touch.
I have noticed increased trafic from the iRobot folks – and would ask them, if they feel comfortable, to comment below on the topic. Hopefully I have captured the story correctly…